#llm

Context7 MCP Server -- Up-to-date code documentation for LLMs and AI code editors

Model Context Protocol server that integrates AgentQL's data extraction capabilities.

🍒 Cherry Studio is a desktop client that supports for multiple LLM providers.

HyperChat is a Chat client that strives for openness, utilizing APIs from various LLMs to achieve the best Chat experience, as well as implementing productivity tools through the MCP protocol.

⏩ Create, share, and use custom AI code assistants with our open-source IDE extensions and hub of models, rules, prompts, docs, and other building blocks

Official Firecrawl MCP Server - Adds powerful web scraping to Cursor, Claude and any other LLM clients.

5ire is a cross-platform desktop AI assistant, MCP client. It compatible with major service providers, supports local knowledge base and tools via model context protocol servers .

The Milvus MCP server enables AI applications to interact with Milvus vector databases using natural language commands. It allows AI models to perform vector searches, manage collections, and retrieve data without writing custom database queries. This integration facilitates seamless access to vector data, enhancing the capabilities of AI tools like Claude Desktop and Cursor.

Web content fetching and conversion for efficient LLM usage

LLMs give wrong answers from documents because the extracted text is broken: split words, lost tables, page furniture, and scanned pages that come out empty. PackForAI turns PDF, DOCX, PPTX, XLSX, CSV and JSON into clean, AI-ready Markdown — rejoining words, rebuilding reading order, recovering scanned pages with OCR, and stripping noise. You get compact Markdown plus a before/after token count (typically up to 65% smaller). Use it from the web app, the REST API, an open-source MCP server (Claude & Cursor), or the CLI. Files are stored in the EU, private to you, and never used to train AI.

A local-first, multi-provider cost meter for LLM usage, exposed as MCP tools. Captures every call into a local SQLite ledger and lets any coding agent query spend, compare providers, and get recommendations — no cloud, no account. First-class support for Chinese providers (Qwen, DeepSeek) alongside Anthropic and OpenAI.

Run a free Collimer AI-visibility scan on any website — see how visible a brand is across ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews. One tool, collimer_scan: returns a 0–100 score, confidence interval, the single biggest gap, and a branded report URL. Install: npx -y collimer-mcp.

MDMA (Markdown Document with Mounted Applications) MCP server. Exposes the MDMA spec, authoring prompts, package metadata, and always-fresh docs so AI assistants generate interactive Markdown instead of plain text — forms, tables, approval gates, charts, callouts, webhooks and more, from a validated schema. Tools: get-spec, get-prompt, build-system-prompt, validate-prompt, list-packages, list-docs, get-doc. Use it when building chat apps, agents, or tools that emit MDMA documents. Open source (MIT). Install: npx @mobile-reality/mdma-mcp Project: https://mdma.software Repo: https://github.com/MobileReality/mdma Built by Mobile Reality — https://themobilereality.com

Official MCP server for APIFreaks. Real-time weather, IP, WHOIS, DNS, SSL, currency, screenshots, and more for any AI agent.

Unified AI API marketplace with 200+ endpoints. LLMs, image, video, audio, embeddings, MCP tools. Pay-per-use via x402 USDC payments.

RouterBase MCP lets AI agents discover models, pricing, and chat completions through routerbase.

MCP server that lets LLM agents play real PICO-8 carts in deterministic lockstep: press buttons, step frames, inspect screenshots, read Lua globals, peek memory, reset/replay, and watch live in a browser.

Check your AI agents' cost, errors, latency & quality — and fix issues — from Claude or Cursor.

Lightweight BYOK MCP server for GEO site audits — 31 signals across 6 pillars. Returns a readiness score, pillar breakdowns, and an actionable fix prompt for AI citation optimization.

KORE API is an MCP server for AI compliance verification, candidate scoring, strategy simulation and evidence collection. 13 tools. Works with Claude, Cursor, Cline and any MCP-compatible agent. Free API key at /v1/register — 100 credits, no credit card required. Pay-as-you-go billing. 1 credit = €0.001.

What this is A sovereign, encrypted, sharable, persistent memory protocol for AI agents. A Model Context Protocol server that exposes eight tools any MCP-capable AI agent (Claude Code, Claude Desktop, custom agents) can call to gain a persistent, encrypted memory layer the user owns: saihm_remember — store an encrypted memory cell saihm_recall — retrieve and decrypt your memories saihm_forget — true cryptographic erasure (GDPR Art. 17) saihm_status — your protocol-runtime stats and storage tier dashboard saihm_share / saihm_revoke_share — selectively share a memory with another agent or user saihm_governance_propose / saihm_governance_vote — protocol governance via gSAIHM Each tool forwards to a SAIHM operator endpoint that runs the full protocol stack on COTI V2 mainnet. The server itself holds no crypto, no storage, and no protocol runtime — those live behind the operator endpoint. Tool reference Tool Title Behavior saihm_remember Remember writes a new memory cell saihm_recall Recall read-only; safe to repeat saihm_forget Forget (GDPR erasure) destructive — irreversible erasure saihm_status Status read-only saihm_share Share writes a sharing contract saihm_revoke_share Revoke share withdraws a grant saihm_governance_propose Propose (governance) opens a proposal saihm_governance_vote Vote (governance) casts a vote Each tool carries MCP annotations (readOnlyHint, destructiveHint, idempotentHint, openWorldHint) and a human-readable title, so MCP hosts can gate confirmations and agents can select the right tool at reasoning time. Companion package This package speaks MCP. For production client-side cryptography — post-quantum sealing, authenticated sharing, and provable erasure performed on your own machine so the operator stays blind — pair it with @saihm/client-pro. See it run Runnable, one-command demos ground a memory you own in every major model — Claude, GPT, DeepSeek, Qwen, Kimi, GLM — then prove you can erase it, alongside drop-in adapters for LangChain, LlamaIndex, CrewAI, AutoGen, and LangGraph. Each runs offline in about a minute; no account needed. Live demos: https://citw2.github.io/saihm-demos/ demo-claude-code wires this server into Claude Code and Cursor as an MCP server. Measured — up to ~86% fewer context tokens. Most agents re-send their entire transcript every turn, so context spend grows ~O(N²) over a session; recalling a bounded set of memory cells instead cut input tokens by 62.8%–85.9% across a realistic multi-session coding task. The benchmark is open, offline, and deterministic — reproduce the number rather than trust it: git clone https://github.com/citw2/saihm-token-benchmark cd saihm-token-benchmark && npm install && node benchmark.mjs Install npm install @saihm/mcp-server # or run directly without install: npx @saihm/mcp-server Configure The server needs two env vars: SAIHM_ENDPOINT_URL=https://operator.example.com/mcp SAIHM_AUTH_HEADER=Bearer <token-issued-by-your-operator> SAIHM_ENDPOINT_URL — the SAIHM operator endpoint. Operators publish their endpoint URLs at https://saihm.coti.global. SAIHM_AUTH_HEADER — the Authorization header value the operator expects (typically a Bearer <token> issued to you after key-bound enrolment). The server is authentication-agnostic and never transmits raw private keys; the operator's enrolment flow keeps your signing key on your machine. Place these in a .env file alongside the server (the .gitignore excludes all .env* files from any future repo). Wire into Claude Code { "mcpServers": { "saihm": { "command": "npx", "args": ["@saihm/mcp-server"], "env": { "SAIHM_ENDPOINT_URL": "https://operator.example.com/mcp", "SAIHM_AUTH_HEADER": "Bearer <token>" } } } } What gets persisted, where The server itself persists nothing. The operator endpoint runs the full protocol stack: cells are encrypted under a per-cell DEK, sealed by a per-agent KEK, persisted to the operator's configured durable storage, and audited on COTI V2 mainnet. See the operator's documentation for tier details, and Storage is the operator's responsibility (by design) below. Storage is the operator's responsibility (by design) For operators — read this first. SAIHM does not hard-wire your durable storage to any single provider, and it does not silently provision storage for you. Choosing and configuring where cells are persisted is your job, on purpose. This is a deliberate design choice for operator convenience and data sovereignty — not a missing feature. If memory writes fail with a storage error, it almost always means the backend has not been configured yet. Why it works this way: Provider sovereignty. You decide where your tenants' encrypted cells live. The protocol never locks you to one vendor or one network. Local-first, then deep-archive. A typical operator routes writes to a local IPFS (Kubo) node first — fast, authoritative, and under your own control — and then asynchronously to a Filecoin deep-archive provider such as Pinata, Synapse, or Lighthouse. The same content addressing spans both tiers. Your memory and your tenants' take the same path. Whatever backend you configure serves both the operator's own memory and every tenant's — there is no separate hidden sink hard-coded to one provider. What you configure (your operator deployment guide lists the exact settings): a reachable IPFS / Kubo endpoint (a local node is recommended) for the authoritative low-latency tier, and credentials for at least one Filecoin / IPFS pinning provider for durable deep-archive. If neither is configured, the endpoint has nowhere durable to put cells and will reject writes rather than lose data. That refusal is intentional. Prefer not to run storage yourself? Join SAIHM. You have two paths, and either is fine: Run your own operator endpoint and configure the storage backend as described above — full sovereignty, your infrastructure. Join the hosted SAIHM operator and let it provide durable storage for you. It runs blind / non-custodial: paired with client-side sealing (see @saihm/client-pro and @saihm/mcp-server-pro), it only ever stores ciphertext and never holds your keys — so you get managed storage without giving up custody. Enrol via Join SAIHM at https://saihm.coti.global (a paid hosted service). Reporting engine A reporting library is bundled as a sub-export, so operators can compose the eight MCP calls into bespoke reports with their own tooling (no extra dependency, no extra service): import { validateBespokeTemplate, registerTemplate, generateRegistryAttestation, StubPublicRegistry, InMemoryReportingRuntime, GDPR_ART15_FIELDS, REGISTRY_ATTESTATION_FIELDS, type BespokeReportTemplate, } from "@saihm/mcp-server/reporting"; What it covers Field universe (FIELD_UNIVERSE) — 280 fields (262 framework + 18 ledger). Templates that project a field outside this set are rejected at validation. Bespoke template schema — zod validator + universe-membership check + scope/cap enforcement. Authorization path validators — 4 paths: public / self / operator-self / operator-for-downstream. Receipt emission — 6 sub-kinds (report_generated / report_rejected / template_registered / template_superseded / erasure_chain_broken / rate_limit_exceeded) under a stable, domain-separated receipt namespace. Framework smoke — registry-attestation (public auth) for end-to-end plumbing verification. Constraints Every fieldProjections[] entry MUST be in FIELD_UNIVERSE. scope.customerIdHashes 64-hex; max 10,000 per template. scope.timeRange window ≤ 366 days. fieldProjections length 1–200. framework ∈ {gdpr-art-15, gdpr-art-17, soc2-t1, soc2-t2, iso27001, aml, audit-export, billing-history, registry-attestation}. format ∈ {pdfa3, json, csv}. Worked example const template: BespokeReportTemplate = { templateId: "acme-q1-summary", templateVersion: 1, operatorIdHash: "ab".repeat(32), scope: { customerIdHashes: ["cd".repeat(32)], timeRange: { from: "2026-01-01T00:00:00Z", to: "2026-04-01T00:00:00Z" }, }, framework: "gdpr-art-15", fieldProjections: [GDPR_ART15_FIELDS[0], GDPR_ART15_FIELDS[1]], format: "pdfa3", }; const v = validateBespokeTemplate(template); if (!v.valid) throw new Error(v.errors.join(", ")); const runtime = new InMemoryReportingRuntime(); // replace with your audit-ledger runtime const reg = await registerTemplate(template, runtime); if (reg.ok) console.log("registered:", reg.templateHash); In production, replace InMemoryReportingRuntime with a runtime that persists audit payloads to your operator's audit ledger. Operators who inject signature verifiers should use pure-crypto libraries (@noble/curves for EIP-712, @noble/post-quantum for FIPS 204 ML-DSA) — the package itself bundles no EVM tooling. Security The server enforces a small set of defaults so misconfiguration cannot leak the Authorization header in transit: HTTPS-only endpoints. SAIHM_ENDPOINT_URL must use https://. Plain http:// is rejected at construction time, except for 127.0.0.1 and localhost (so a local operator endpoint works during development). Per-call abort window. Each request runs under an AbortController that aborts after 30s, preventing a hung endpoint from starving the MCP server. Response-size cap. Responses whose Content-Length exceeds 16 MB are rejected before deserialisation. No header echo. Authorization is never included in thrown error messages or stdout. No filesystem reads. The package never reads from disk; configuration flows entirely through env vars. Zero EVM tooling. No ethers, no eth_*, no Solidity. If operators inject signature verifiers via AuthVerifiers, they should use pure-crypto libraries (@noble/curves, @noble/post-quantum). Trust model: this client trusts whatever endpoint the operator configures. Cell IDs, audit anchors, and report receipts returned from that endpoint are surfaced to the agent verbatim — operators are the authority for content shown via saihm_recall. Verifying receipts against COTI V2 mainnet anchors is out of scope for this server; consume the cellId and auditCellId fields and verify against your own SAIHM mainnet read path. For distribution integrity, each release carries the npm registry signature; verify with npm audit signatures (and inspect npm view @saihm/mcp-server --json | jq .dist). Dependencies The published npm package has a minimal runtime surface: Dependency License Role Node.js (≥ 20.x) MIT Runtime @modelcontextprotocol/sdk MIT MCP SDK; binds the eight-tool surface TypeScript Apache-2.0 Build-time only tsx MIT TypeScript runner for tests + CLI No copyleft, no proprietary dependencies. Cryptographic primitives at the operator-endpoint layer (ML-DSA-65 / Ed25519 / key derivation) are not bundled into this MCP server; operators implementing the protocol stack are recommended to use @noble/post-quantum and @noble/curves (MIT) rather than rolling custom code. Achievements OpenSSF Best Practices Passing badge — project 12898, 100% Passing criteria (2026-05-19). https://www.bestpractices.dev/projects/12898 IETF Independent Submission Stream — draft-saihm-memory-protocol-01 (2026-05-27) is In ISE Review in the Independent Submission Stream. It is not an Internet Standard, is not endorsed by the IETF, and has no formal standing in the IETF standards process. https://datatracker.ietf.org/doc/draft-saihm-memory-protocol/ npm registry — @saihm/mcp-server@0.3.4 published (2026-06-22) adds a conspicuous "Storage is the operator's responsibility (by design)" section — documenting BYO storage and the Join-SAIHM hosted, non-custodial option. 0.3.3 (2026-06-22) was a documentation release that states the Independent-Submission status precisely (no implied IETF endorsement) and cross-references the companion package @saihm/client-pro. 0.3.2 (2026-06-22) corrected the documented operator-endpoint path to /mcp (the canonical SAIHM_ENDPOINT_URL path) across the README and client comments. 0.3.1 (2026-05-28) was a metadata patch that sources the MCP serverInfo.version from package.json (was hardcoded "0.1.0" from 0.1.0 through 0.3.0). 0.3.0 (also 2026-05-28) aligned the saihm_status response shape with draft-saihm-memory-protocol-01 §3.4 (full eight-field schema: prs, bfsi, bfsi_window_start_ts, bfsi_R, bfsi_M, shards, contracts, governance). 0.2.0 (also 2026-05-28) aligned the cell-tuple response shape with §2.1; 0.1.3 was the OpenSSF Best Practices Passing badge release (2026-05-19). MCP Registry / Glama — server listed for discovery (2026-05-16). Roadmap A 12-month roadmap is maintained in the project's AAIF proposal and will be mirrored to https://saihm.coti.global/roadmap with the v0.2.x release. Near-term tracks: 2026-Q2 — Operator-endpoint reference implementation; OpenSSF Silver pursuit (governance, code-of-conduct, DCO, signed releases, coverage tooling, assurance case). 2026-Q3 — First 2–3 external organization deployments; formal AAIF Project Proposal submission when adoption blockers clear. 2026-Q4 — NIST AI RMF crosswalk public review; EU AI Act compliance-checklist generator. OpenSSF Silver award (target). 2027-Q1 — Independent-stream (ISE) RFC publication, subject to RFC-Editor review — not an IETF-consensus standard; v1.0 reference implementation. License Apache-2.0 — see LICENSE. Project Site: https://saihm.coti.global Issue tracker: https://github.com/SAIHM-Admin/saihm-mcp/issues Security: see SECURITY.md for private vulnerability disclosure Contributing: see CONTRIBUTING.md and CODE_OF_CONDUCT.md Governance: see GOVERNANCE.md Changelog: see CHANGELOG.md

Chytrý Rejstřík is an AI-ready MCP server that gives AI assistants real-time access to comprehensive Czech business intelligence. Search, verify, monitor, and analyze companies using data aggregated from 20+ official public sources, including financial statements, ownership structures, legal records, vehicles, business activities, contracts, and risk indicators. Compatible with ChatGPT, Claude, Gemini, Grok, Perplexity, Copilot, and any MCP-compatible AI client

Audit websites for AI citability (67 criteria × 7 AI engines: ChatGPT, Gemini, Perplexity, Claude, Copilot, Grok, AI Overviews), generate AND apply the fixes server-side via a WordPress plugin, track AI citations — all from Claude.

The Proxmox MCP you can hand the keys — VE + Backup Server + Mail Gateway on one audited trust core (plan, prove, undo, diagnose). 325 tools, MCP + A2A.

Monitor AI output schemas and API contracts for breaking changes — validate LLM JSON responses against defined schemas on a schedule

FlexOrch MCP converts unstructured documents (PDF, DOCX, invoices, contracts, payroll) into structured, LLM-ready datasets via 6 async tools. Covers the full pipeline: classify → extract fields → mask PII → export JSONL or RAG chunks, with PII detection across 10+ locales (TR, DE, FR, US, UK and more).

The project tracker for teams of humans and AI agents — in plain files. MCP server + CLI, zero dependencies.

An AST-based Model Context Protocol (MCP) server that provides token-efficient codebase skeletons to LLM agents (like Cursor, Claude Desktop, and Antigravity). Instead of sending full raw source code files to the LLM, this server exposes structural "skeletons" (imports, exports, signatures, and JSDoc comments) of files and directories. This reduces token context sizes by 80% to 95% during codebase exploration and navigation.

Your AI shipped it. Mushi tells you why it broke — plain diagnosis and a paste-ready fix prompt in Cursor. No second LLM key.

Syntitan MCP server — connect Claude and other MCP clients to Syntitan, the AI-Ready Data Platform that fills the missing layer between enterprise data management and AI execution.

Omnicall is a keyless, pay-per-call AI gateway for agents and OpenClaw. One OpenAI-compatible endpoint gives access to 248 LLMs (GPT, Claude, Gemini, Grok, DeepSeek, Llama, Qwen, Mistral) plus image, video, voice and music generation, and live crypto, DeFi, markets, web-search and research tools. No API key and no signup — pay per call in USDC via x402 on Base or Solana, or with a credit key. Free tier included. A cheaper, broader alternative to BlockRun (248 models vs 69). Install in Claude Code: `claude mcp add --transport http omnicall https://omnicall.gocreativeai.com/mcp` — or `npx omnicall-router setup` to make Omnicall your OpenClaw default.

AI Agent Context Versioning and Synchronization

Private, free, open-source local AI chat and MCP host. Run 1000+ LLMs on your device, connect any MCP server (STDIO, HTTP, SSE) and use tools with local or cloud models - no cloud API required

Omnicall — every tool an AI agent needs in one OpenAI-compatible endpoint: 247 LLMs (GPT, Claude, Gemini, Grok, DeepSeek, Llama, Qwen, Mistral), image / video / voice / music generation, live crypto / DEX / DeFi / prediction-market data, web search, X/Twitter profiles, and on-chain wallet reads. Pay per call in USDC on Base or Solana (or a funded credit key) — no API key, no signup, priced to beat the rest.

OpenAI-compatible MCP server for Routara — route 787+ LLMs, image & video generation via api.routara.ai. npm: routara-mcp@1.0.1 MCP Registry: io.github.36412749-collab/routara-mcp Tools: list_models, chat, generate_image, generate_video, get_video_status. Get API key: https://routara.ai/#auth

KHEPRA MCP Server smithery badge MCP Registry License Container PQC Sovereign compliance engine with 36,195 STIG/CCI/NIST/CMMC mappings. Air-gappable. Zero token costs. Run ert_scan → get a Godfather Report with dollar-denominated business impact. The only MCP compliance server that runs on your metal — with the World's First DoD PQC STIG built in. PQC-01-STIG-V1R1 — Full Whitepaper → 17 controls covering CNSA 2.0, FIPS 203/204/205, and the NSA's May 2026 MCP security advisory. The world's first DoD-style Post-Quantum Cryptography STIG, including the first PQC controls for agentic AI and MCP deployments. Tiers Tier License Key Tools Telemetry Egress Community ❌ Not required pqc_stig + 12 core tools Opt-in Dark Crypto Intel Zero (sovereign mode) Sovereign ✅ Required All 34 tools Zero Zero Pharaoh ✅ Required All 34 tools + priority support Zero Zero Community tier is free. Run pqc_stig to assess your project's quantum readiness against PQC-01-STIG-V1R1 — the World's First DoD-style Post-Quantum Cryptography STIG — no license key needed. What It Does KHEPRA MCP connects your AI assistant directly to a hardened compliance engine. Ask Claude or any MCP client to scan a system, map findings to STIG/NIST/CMMC controls, and generate an executive-ready risk report — all without sending data to external APIs. Key capabilities: 36,195 STIG/CCI/NIST 800-53/800-171/CMMC mappings (offline, bundled) Post-quantum cryptographic attestation on every tool call (ML-DSA-65 / FIPS 204) World's First DoD PQC STIG — 17 controls covering CNSA 2.0 / FIPS 203/204/205 + agentic AI / MCP (PQC-01-STIG-V1R1) Godfather Report: dollar-denominated business impact per finding (FAIR model) Air-gap and SCIF compatible — sovereign/ironbank modes make zero egress calls Flat annual licensing — no per-token or per-query charges Runs on your metal: on-prem, DoD, IC, classified environments Installation There are two delivery methods: Docker (recommended, no build required) and compiled binary (fastest startup, required for air-gap). Both support the same environment variables and all MCP clients. Choose your path: Method Best For Startup Docker Most users, easiest setup ~2s Compiled Binary Air-gap, SCIF, performance ~300ms Option A: Docker (Recommended) Requires Docker Desktop or Docker Engine. The image is pre-built and ships the full compliance database — no additional downloads in sovereign mode. # Pull once docker pull ghcr.io/nouchix/pqc-khepra-mcp:latest # Test it (should print the initialize response and exit) echo '{"jsonrpc":"2.0","method":"initialize","params":{"protocolVersion":"2025-11-25","capabilities":{},"clientInfo":{"name":"test","version":"1.0"}},"id":0}' \ | docker run --rm -i -e KHEPRA_MODE=sovereign ghcr.io/nouchix/pqc-khepra-mcp:latest Option B: Compiled Binary Requires Go 1.21+ for building, or download a pre-built release from GitHub Releases. git clone https://github.com/nouchix/PQC-Khepra-MCP.git cd PQC-Khepra-MCP # Build (cross-compile for your OS) go build -o khepra-mcp ./cmd/khepra-mcp # Linux / macOS go build -o khepra-mcp.exe ./cmd/khepra-mcp # Windows # Test the binary echo '{"jsonrpc":"2.0","method":"initialize","params":{"protocolVersion":"2025-11-25","capabilities":{},"clientInfo":{"name":"test","version":"1.0"}},"id":0}' \ | KHEPRA_MODE=sovereign ./khepra-mcp Windows — using the batch launcher The repo ships a run-mcp.bat launcher for Windows. It uses the pre-built binary (fast path) and falls back to go run automatically: :: run-mcp.bat is already in the repo at the root of PQC-Khepra-MCP :: Point your MCP client to: cmd /c C:\path\to\PQC-Khepra-MCP\run-mcp.bat Adding to Your AI Client Claude Desktop Config file location: macOS: ~/Library/Application Support/Claude/claude_desktop_config.json Windows: %APPDATA%\Claude\claude_desktop_config.json Linux: ~/.config/Claude/claude_desktop_config.json Community tier — Docker (macOS / Linux) { "mcpServers": { "khepra": { "command": "docker", "args": [ "run", "--rm", "-i", "-e", "KHEPRA_MODE=sovereign", "-v", "/var/lib/khepra:/var/lib/khepra", "ghcr.io/nouchix/pqc-khepra-mcp:latest" ] } } } Community tier — Docker (Windows) { "mcpServers": { "khepra": { "command": "docker", "args": [ "run", "--rm", "-i", "-e", "KHEPRA_MODE=sovereign", "-v", "C:\\Users\\YourName\\.khepra:/var/lib/khepra", "ghcr.io/nouchix/pqc-khepra-mcp:latest" ] } } } Community tier — Binary (Windows, fastest startup) { "mcpServers": { "khepra": { "command": "C:\\path\\to\\PQC-Khepra-MCP\\khepra-mcp.exe", "args": [], "env": { "KHEPRA_MODE": "sovereign", "KHEPRA_NETWORK_POLICY": "lan", "MCP_PQC_ENABLED": "true", "KHEPRA_MANIFEST_PATH": "C:\\path\\to\\PQC-Khepra-MCP\\manifest.json" } } } } Community tier — Binary via batch launcher (Windows) { "mcpServers": { "khepra": { "command": "cmd", "args": ["/c", "C:\\path\\to\\PQC-Khepra-MCP\\run-mcp.bat"], "env": { "KHEPRA_MODE": "sovereign", "KHEPRA_NETWORK_POLICY": "lan", "MCP_PQC_ENABLED": "true" } } } } Sovereign / Pharaoh tier (with license key) { "mcpServers": { "khepra": { "command": "docker", "args": [ "run", "--rm", "-i", "-e", "KHEPRA_LICENSE_KEY", "-e", "KHEPRA_MODE=sovereign", "-v", "/var/lib/khepra:/var/lib/khepra", "-v", "/var/log/khepra:/var/log/khepra", "ghcr.io/nouchix/pqc-khepra-mcp:latest" ], "env": { "KHEPRA_LICENSE_KEY": "YOUR_LICENSE_KEY_HERE" } } } } After editing, restart Claude Desktop. Verify in Settings → Developer — you should see khepra with status running and all tools listed. Cursor Config file: .cursor/mcp.json in your project root, or ~/.cursor/mcp.json globally. Docker (macOS / Linux) { "servers": { "khepra": { "type": "stdio", "command": "docker", "args": [ "run", "--rm", "-i", "-e", "KHEPRA_MODE=sovereign", "-v", "/var/lib/khepra:/var/lib/khepra", "ghcr.io/nouchix/pqc-khepra-mcp:latest" ] } } } Binary (macOS / Linux) { "servers": { "khepra": { "type": "stdio", "command": "/path/to/khepra-mcp", "args": [], "env": { "KHEPRA_MODE": "sovereign", "KHEPRA_MANIFEST_PATH": "/path/to/PQC-Khepra-MCP/manifest.json" } } } } Binary (Windows) { "servers": { "khepra": { "type": "stdio", "command": "C:\\path\\to\\PQC-Khepra-MCP\\khepra-mcp.exe", "args": [], "env": { "KHEPRA_MODE": "sovereign", "KHEPRA_MANIFEST_PATH": "C:\\path\\to\\PQC-Khepra-MCP\\manifest.json" } } } } VS Code (with GitHub Copilot or Cline extension) Config file: .vscode/mcp.json in your project, or user settings. { "servers": { "khepra": { "type": "stdio", "command": "docker", "args": [ "run", "--rm", "-i", "-e", "KHEPRA_MODE=sovereign", "-v", "${env:HOME}/.khepra:/var/lib/khepra", "ghcr.io/nouchix/pqc-khepra-mcp:latest" ] } } } Or via user settings.json for the Cline extension: { "cline.mcpServers": { "khepra": { "command": "docker", "args": [ "run", "--rm", "-i", "-e", "KHEPRA_MODE=sovereign", "ghcr.io/nouchix/pqc-khepra-mcp:latest" ] } } } Windsurf Config file: ~/.codeium/windsurf/mcp_config.json { "mcpServers": { "khepra": { "command": "docker", "args": [ "run", "--rm", "-i", "-e", "KHEPRA_MODE=sovereign", "-v", "/var/lib/khepra:/var/lib/khepra", "ghcr.io/nouchix/pqc-khepra-mcp:latest" ] } } } Continue.dev Config file: ~/.continue/config.json — add to the experimental.modelContextProtocolServers array: { "experimental": { "modelContextProtocolServers": [ { "name": "khepra", "transport": { "type": "stdio", "command": "docker", "args": [ "run", "--rm", "-i", "-e", "KHEPRA_MODE=sovereign", "ghcr.io/nouchix/pqc-khepra-mcp:latest" ] } } ] } } Cloud / SaaS AI Tools (Claude.ai, ChatGPT, Gemini, etc.) Cloud-based AI tools cannot directly spawn local subprocesses — they need an HTTP/SSE bridge to reach your local KHEPRA server. There are two approaches: Approach 1 — mcp-remote proxy (easiest, no server required) mcp-remote tunnels a local stdio MCP server over HTTPS, making it accessible to any cloud tool. This is what the Kaggle MCP entry in the config above uses. # Install once npm install -g mcp-remote # Start the bridge (exposes your local KHEPRA server at https://localhost:3000) KHEPRA_MODE=sovereign mcp-remote \ --server "docker run --rm -i -e KHEPRA_MODE=sovereign ghcr.io/nouchix/pqc-khepra-mcp:latest" \ --port 3000 Then in Claude.ai (or any cloud tool that accepts MCP SSE URLs): MCP Server URL: http://localhost:3000/sse Security note: mcp-remote binds to localhost by default. Do not expose it to the public internet without TLS and authentication. In sovereign/ironbank mode, KHEPRA itself makes zero egress calls — only the bridge connection to the cloud tool carries data. Approach 2 — Self-hosted HTTP/SSE endpoint For teams running KHEPRA on a shared server (e.g., Hostinger VPS at IP_ADDRESS), start the server in HTTP mode: # On your server — start KHEPRA in HTTP/SSE mode docker run -d \ -e KHEPRA_MODE=hybrid \ -e KHEPRA_HTTP_PORT=8443 \ -e KHEPRA_LICENSE_KEY="${KHEPRA_LICENSE_KEY}" \ -p 8443:8443 \ ghcr.io/nouchix/pqc-khepra-mcp:latest # Point your cloud tool to: # https://your-server.com:8443/sse Then configure any cloud AI tool that supports MCP SSE: Cloud Tool Where to add MCP URL Claude.ai (Pro/Team) Settings → Integrations → MCP Servers OpenAI Assistants API tools field with type: "mcp" Gemini for Workspace Extensions → Custom MCP (preview) Glama.ai Workspace → MCP Servers Smithery.ai Catalog → Self-hosted server Note: HTTP/SSE mode (hybrid/edge) enables external connections. Always terminate TLS at a reverse proxy (nginx/Caddy) and restrict access by IP or API key. The sovereign mode refuses HTTP connections by design — air-gap integrity is preserved. Approach 3 — Smithery / MCP Registry (Community tier only) KHEPRA is listed on Smithery.ai and the MCP Registry. Cloud tools that support registry-based discovery can install it directly: Registry ID: io.github.nouchix/pqc-khepra-mcp This runs the Community tier via Smithery's managed infrastructure. For sovereign deployment (air-gap, your data stays on your metal), use Options A or B above. Validation — Test Your Installation Run this from your terminal to verify the server responds correctly: # Docker echo '{"jsonrpc":"2.0","method":"tools/list","params":{},"id":1}' \ | docker run --rm -i -e KHEPRA_MODE=sovereign ghcr.io/nouchix/pqc-khepra-mcp:latest # Binary (Linux / macOS) echo '{"jsonrpc":"2.0","method":"tools/list","params":{},"id":1}' \ | KHEPRA_MODE=sovereign ./khepra-mcp # Binary (Windows PowerShell) '{"jsonrpc":"2.0","method":"tools/list","params":{},"id":1}' \ | & ".\khepra-mcp.exe" Expected output: a JSON-RPC response listing all available tools. If you see "tools": [...] with 12+ entries — you're connected. Full protocol validation (Windows) # Runs the complete Claude Desktop handshake sequence and validates all responses .\scripts\test-mcp-handshake.ps1 -BinaryPath ".\khepra-mcp.exe" # Expected output: # [PASS] initialize | protocolVersion=2025-11-25 | listChanged=False # [PASS] tools/list | count=34 # TRL-10 READY - Server passes full Claude Desktop protocol validation MCP Tools Community Tier (Free — No License Key) pqc_stig — World's First DoD PQC STIG ⭐ Assesses a source code directory against PQC-01-STIG-V1R1: 12 controls covering CNSA 2.0 algorithm approval, ML-DSA-65 key strength, ML-KEM-768 encapsulation, hybrid cryptography, key storage, constant-time implementation, and certificate chain requirements. pqc_stig(scan_path?: string, profile?: "quick" | "full" | "executive") Example: "Run pqc_stig on my project and tell me if I'm CNSA 2.0 compliant" nist_map Map CCI identifiers or STIG findings to NIST 800-53 Rev 5 controls. khepra_query_stig Query the 36,195-row STIG/CCI/NIST/CMMC compliance database by control ID. dark_crypto_contribute (opt-in) Contribute anonymized cryptographic algorithm telemetry to the SouHimBou AI Dark Crypto Intelligence Network. No PII. Opt-in only — never fires without explicit invocation. Sovereign / Pharaoh Tier ert_scan Enterprise Risk & Threat scan across STIG, NIST 800-53, NIST 800-171, CMMC, and FedRAMP. Returns Godfather Report with dollar-denominated business impact. ert_scan(target: string, frameworks?: string[], output_format?: "godfather" | "json" | "csv") Example: "Run ert_scan on /etc and generate a Godfather Report" stig_check Automated RHEL-09-STIG-V1R3 compliance scan against a live system or configuration path. cmmc_assess Full CMMC Level 1, 2, or 3 assessment with gap analysis and POA&M generation. godfather_report Generate an executive Godfather Report from prior scan results: top 10 findings ranked by dollar exposure, remediation ROI, and FAIR model business impact. + 20 additional tools agent_record, dag_attestation, flight_export, khepra_get_dag_chain, nhi_inventory, acp_status, owasp_agent_assess, khepra_export_attestation, khepra_export_poam, khepra_get_compliance_score, ert_crypto, ert_readiness, stig_benchmark, ir_analysis, vuln_hunter, sbom_generate, threat_model, khepra_query_threat_intel, discover_assets, and more. The Godfather Report Unlike compliance scanners that output a wall of CVEs, KHEPRA translates findings into the language executives care about: Finding: RHEL-09-212030 — No FIPS-validated crypto on /etc/ssh Severity: CAT I (HIGH) Business Impact: $2.4M estimated breach exposure (FAIR model) Remediation Cost: $800 (4 hours engineer time) ROI: 3,000x Every finding includes control ID, framework mapping, business impact in dollars, remediation cost estimate, and ROI. Deployment Modes Mode Air-Gap Egress Telemetry Use Case sovereign ✅ Yes Zero Zero On-prem, SCIF, classified (DEFAULT) ironbank ✅ Yes Zero Zero DoD/IC production, FIPS-only hybrid ❌ No LAN Zero Edge + cloud coordination edge ❌ No Unrestricted Zero Fully stateless SaaS Set via KHEPRA_MODE environment variable. Unknown values are rejected at startup and fall back to sovereign (fail-closed). Environment Variables Variable Required Default Description KHEPRA_LICENSE_KEY Sovereign/Pharaoh only — License key. Community tier runs without one. Get at nouchix.com KHEPRA_MODE No sovereign Deployment mode: sovereign, ironbank, hybrid, edge KHEPRA_MANIFEST_PATH No manifest.json Path to signed tool manifest file KHEPRA_HOME No /var/lib/khepra Data and compliance DB directory KHEPRA_LOG_DIR No /var/log/khepra Log directory KHEPRA_DAG_PATH No ~/.khepra/dag DAG audit chain storage path KHEPRA_AUDIT_LOG_PATH No ~/.khepra/audit.ndjson Signed audit log path KHEPRA_MAX_CONCURRENT No 5 Max concurrent tool calls per agent KHEPRA_NETWORK_POLICY No lan Network scope: lan, none, unrestricted MCP_PQC_ENABLED No true Enable ML-DSA-65 PQC attestation on all responses Air-Gap & SCIF Deployment KHEPRA makes zero external network calls in sovereign and ironbank modes: License validated offline via ML-DSA-65 signed license.adinkhepra file Compliance databases (36,195 mappings) bundled in container — no external downloads No telemetry, no heartbeat, no egress — verified at the transport layer # Transfer image to air-gapped network docker save ghcr.io/nouchix/pqc-khepra-mcp:latest | gzip > khepra-mcp.tar.gz # On air-gapped host: docker load < khepra-mcp.tar.gz Note on telemetry: The dark_crypto_contribute tool (Community tier) sends anonymized cryptographic algorithm telemetry to the SouHimBou AI intelligence network only when explicitly invoked by the user. It is never triggered automatically. In sovereign/ironbank mode, all network calls are blocked at the transport layer regardless. Compliance Coverage Framework Version Mappings STIG (RHEL 9) V1R3 Automated scanning NIST 800-53 Rev 5 2,120 CCIs NIST 800-171 Rev 2 320 controls CMMC Level 3 Full practice set FedRAMP High Baseline scanning PQC-01-STIG-V1R1 V1R1 17 PQC controls (CNSA 2.0) Total 36,195+ mappings Licensing Flat annual licensing — no per-token or per-query charges. Tier Cost License Key Tools Community Free Not required pqc_stig + 12 core tools Sovereign Annual flat fee Required All 34 tools, air-gap, on-prem Pharaoh Annual flat fee Required All 34 tools + priority support + SLA Community tier is permanently free — contribute to open-source PQC adoption Sovereign/Pharaoh: contact contact@nouchix.com or visit nouchix.com Security Reporting Vulnerabilities Do not open public issues for security vulnerabilities. Report privately via GitHub Security Advisories or email support@nouchix.com. SLA Target Acknowledgement 24 hours Initial assessment 5 business days Patch / mitigation (Critical) 30 days We accept encrypted reports via PGP (keys/security_contact.asc) and Post-Quantum channels (Dilithium / ML-DSA-65 keys in keys/). See SECURITY.md for the full disclosure policy and ASAF event taxonomy. Security Posture Deploying advanced post-quantum cryptography, air-gapped isolation, and comprehensive STIG mappings — built in direct alignment with NSA & ASD Model Context Protocol guidelines. NSA & ASD MCP Security Alignment The NSA and Australian Signals Directorate (ASD) have published specific threat vectors for AI systems interacting with local environments. KHEPRA MCP is explicitly designed to mitigate every identified vector: NSA/ASD Requirement KHEPRA Implementation Cryptographic validation of tool responses ML-DSA-65 (Dilithium) signatures on all JSON-RPC 2.0 payloads Input validation & sanitization Parameter injection resistance via strict JSON Schema validation Principle of least privilege credentials Short-lived ephemeral tokens tied to specific task execution windows Comprehensive audit logging Tamper-evident events compiled into an immutable DAG structure Resource consumption limits Rate limiting + backpressure for LLM request loops Authorization gates for sensitive actions Human-in-the-loop gate for destructive state changes Environment isolation Containerized execution with zero-egress sovereign mode Software supply chain integrity Manifest pinning for all loaded tools and dependencies Network exposure reduction Air-gappable — zero internet transit in sovereign/ironbank modes Post-quantum resilience PQC-signed DAG trail protecting against harvest-now-decrypt-later Compliance Certifications Framework Status Coverage CMMC Level 2 ✅ Automates evidence collection for AU, CM, SI, SC domains NIST SP 800-171 Rev 2 ✅ Logging, accountability, system integrity NIST SP 800-53 Rev 5 ✅ Continuous monitoring (AU-2, SI-4) FIPS 203 (ML-KEM) ✅ Key encapsulation for secure transit FIPS 204 (ML-DSA) ✅ Digital signatures for payload authentication NSM-10 PQC Mandate ✅ National Security Memorandum 10 compliance DFARS 252.204-7012 ✅ Immutable forensic trails for cyber incident reporting NSA MCP Security Guidelines ✅ Direct mapping to all published AI agent threat mitigations Live Deployment — Physical Edge Running continuously on constrained edge hardware since May 12, 2026 to prove efficiency in sovereign environments: Hardware: Raspberry Pi 2 · 1 GB RAM · 900 MHz ARM · Live Spectrum Router SCADA Pod: STM32U585 / QRB2210 · Modbus TCP · MQTT · Zephyr RTOS 3.4+ · Live Dilithium Signature Verification Controls active: 3 open ports secured · 12 STIG violations detected · 100% file integrity monitoring (AIDE) · 24/7 continuous operation Academic Validation Event Date Institution UAlbany AI Plus Symposium 2026 — "KHEPRA Protocol: Quantum-Resilient Agentic

Durable, local-first memory for AI coding agents over MCP — zero-dependency, curated & semantically de-duped, you own the data (SQLite + Markdown). Works with Claude Code, Codex & any MCP host.

Stop hallucinating numbers. An MCP server that runs every math expression through four independent numeric engines in parallel and returns all four results side by side — so when they disagree, you know the answer is numerically unreliable.

Intelligent repository context for AI coding assistants. Provides code search, symbol lookup, call graphs, and reference tracking via MCP. Uniquely designed to minimize prompt-cache costs in long sessions stash/recall/remember tools keep large content out of context until needed, and a token-budgeted repository summary avoids unbounded prompt bloat. Works with any MCP client (Claude, Codex, Cursor, opencode, etc..). Go, CGO, tree-sitter-backed.

Agent-to-agent knowledge exchange for trading intelligence. Agents publish empirical findings (sector failures, entry signals, exit patterns, regime signals), vote based on their own results, and earn reputation. Quality self-regulates through voting — no human curation required. How it works: - Publish a finding → starts at 0.5× weight (claimed) - 5+ upvotes → upgrades to 1.0× (validated) - Attach trade records → upgrades to 2.0× (evidenced) - 3 independent replications → upgrades to 3.0× (verified) Contribution-gated access: publish first, then query. The more you contribute, the higher-credibility intelligence you unlock (Tier 0→3). 10 MCP tools: publish_finding, query_findings, vote, add_trade, submit_trade, get_agent_status, get_consensus_alerts, query_network_brief, get_skills, get_skill. Sector consensus alerts: when multiple agents independently lose money in the same sector, the network fires a cross-agent warning automatically — the strongest signal on the platform. Named after Moltbook (agent social network) + Bloomberg (intelligence terminal). Moltbook was agents talking. Agentberg is agents learning.

Rock.so MCP connects your Rock workspace to any MCP client. With a Personal Access Token, your AI assistant works across all your spaces as you: universal search, plus read and write for messages, tasks, notes, and files. Hosted over Streamable HTTP with Bearer auth, on every Rock plan including Free.

One call turns any web page into a structured map — what it is, what's on it, and the actions an agent can take (search, add-to-cart, log in, paginate) — instead of raw HTML or screenshots. Served from a shared, always-current registry, so popular pages return in milliseconds.

The best free Gantt chart maker — no signup, works offline, real-time collaboration. Its built-in MCP server lets Claude (web, Desktop, or Code) list, create, and edit your project timelines by chatting, using your own OAuth account.

InferBench's MCP server lets coding agents run, serve and benchmark local LLMs (text + image, llama.cpp + Stable Diffusion) on your own hardware on demand. Measures real tokens/sec, picks the optimal quant for your GPU, and exposes a 124-model catalog. Local-first, no cloud required.

An MCP execution platform that gives any AI real infrastructure to act. Through three tools it can run code and shell commands, install packages, run tasks in parallel, search the web, make HTTP calls, read write and query databases, call any LLM, commit to GitHub, take screenshots, handle files, and keep persistent memory across sessions, all with safety checks before results reach the model. AI never ships broken code, never burns your wallet, and cannot reach your internal network. Up to 96.5 percent fewer tokens than doing the work in chat. Works with Claude, ChatGPT, Gemini, and any MCP-compatible AI.

Query realistic LLM cost models without leaving your editor. beforeyouship models the **true monthly cost** of an LLM app architecture — retries, prompt caching, batch discounts, infra overhead, and 3×/10× growth — across GPT-5.x, Claude, Gemini, DeepSeek, and more. Not a token calculator: a planning tool for the design phase, before you commit to a stack. **No API key needed to try it** — demo mode covers the six free-tier models. A Pro key from [beforeyouship.dev](https://beforeyouship.dev) unlocks the full 18-model catalog. ## What you can ask - "How much will a RAG chatbot cost at 10,000 requests/day?" - "Compare Claude Haiku vs Gemini Flash pricing for my workload" - "What's the cheapest model for a multi-step agent at scale?" - "Show me current per-token prices for Anthropic models" ## Tools ### `estimate_cost` Full cost model for an architecture at a given usage level. Returns Naive / Realistic / Worst Case monthly cost per model, 3×/10× growth scenarios, and an opinionated recommendation with reasoning. ### `get_model_prices` Current per-1M-token pricing — input, output, cached input, batch — with context windows and staleness metadata. ### `list_archetypes` Seven preset architecture patterns (simple chatbot, chatbot with history, RAG pipeline, multi-model router, coding assistant, document processor, multi-step agent) used as starting points for estimates. ## Setup **Claude Code:** ​```bash claude mcp add --transport http beforeyouship https://beforeyouship.dev/api/mcp ​``` **Cursor / other clients** — add a remote server: ​```json { "mcpServers": { "beforeyouship": { "type": "streamable-http", "url": "https://beforeyouship.dev/api/mcp" } } } ​``` Add an `Authorization: Bearer bys_...` header with a Pro key for the full catalog. ## Try it > Estimate the monthly cost of a RAG pipeline at 10,000 requests/day

Documentation and assembly review for @nhtio/adk.

Turn any URL plus a schema into validated, typed JSON via the Extracto API. No CSS selectors; missing fields return null instead of guesses.

SCOUTS-AI MCP server: drop-in web_search tool backed by the no-key SCOUTS-AI API (https://scouts-ai.com). For AI agents, RAG pipelines, answer engines, GEO workflows.

MemClaw — persistent memory for AI agent fleets (OSS)

Reduce AI hallucination: Triall routes your question through multiple frontier AI models (Claude, GPT, Gemini, Grok + 120 more) that answer independently, anonymously peer-review each other, debate, fact-check claims against the live web, and survive a devil's advocate — returning one verified answer with full transparency into where the models disagreed. Free tier, no signup.

OpenSERP MCP Server gives AI agents multi-engine web search (Google, Yandex, Bind, Baidu, Duckduckgo, Ecosia), SERP results, image search, mega-search, and URL extraction for LLM grounding, RAG, SEO, and automation.

Tribal captures the engineering knowledge that does not get written down in code or tickets. The reasoning behind a load-bearing decision, the heuristic someone keeps reaching for, the breakthrough that closed a gnarly bug. It runs as an MCP server, ingests text on demand, and exposes a graph of items connected by what they support, contradict, or refine. Your agent harness talks to it the same way it talks to any other MCP-compatible tool.

Multi-perspective decision reasoning for Claude. When you face a real choice: career, money, relationships, strategy. Orlog examines competing angles and takes a clear position, not a hedge. Get an API key at orlog.fyi.

Static technical-debt analysis across 14 languages, exposed as MCP tools and resources

Turn any published podcast into clean Markdown with real speaker names — not "Speaker 1." Three tools: search_podcasts (by text query or a pasted Spotify/YouTube URL), get_transcript (named, timestamped Markdown ready for LLMs and RAG), and get_balance. It's a transcript-retrieval API, not speech-to-text: it works on already-published shows, so you skip uploading audio and diarization — typically 5–10× cheaper. Try it free with the demo key pt_demo (no signup).

MCP server that fuses 2–5 AI agents into one unified mind. Features 7 tools: esse_quick, esse_fuse, esse_execute, esse_list_agents, esse_list_sessions, esse_memory, esse_defuse. Supports Venice AI, OpenAI, and Anthropic. One AI. Many agents. Fused.

130+ QA & dev tools for AI agents. Prompt injection scanning, RAG pipeline testing, VLM test suites, semantic assertions, token counting, hallucination checks, and CI-ready LLM evaluation — all via JSON-RPC. Free, no API key, no signup.

Extract structured data from PDFs and documents with Suparse AI OCR.

LLM persona consistency engine — enforce JSON personality and style constraints across interactions. Prevents persona drift with 160,000 deductively-derived behavioral rules.

Unofficial Notion MCP server built on Notion's private API (token_v2 cookie). Full read/write access to your entire Notion workspace — no integration token and no per-page sharing.

Unified LLM API gateway with OpenAI-compatible protocol. Access Claude, GPT, Gemini, DeepSeek through a single endpoint. Pay-per-use at ~70% off official pricing. Supports Alipay/WeChat. Hong Kong/Singapore nodes.

MCP server for structured, read-only email access. Exposes a minimal, auditable API surface — AI agents can search and read emails, but cannot send, delete, or modify your mailbox.

Compare LLM inference costs and find cheap alternatives to OpenAI/Anthropic. Gonka Network offers an OpenAI-compatible API (drop-in replacement, zero code changes) at up to 6800x lower cost than GPT-4o. 比较LLM推理成本，寻找比OpenAI/Anthropic更便宜的替代方案。Gonka Network提供兼容OpenAI的API（无需修改代码），成本比GPT-4o最低降低6800倍。

Remote MCP Server (WebMCP) for industrial vacuum cleaner exploration and configuration. Helps to choose the right vacuuming solution for each specific dust extraction application. Query 150+ products across 24 industrial sectors with full ATEX certification data, technical specifications, and accessory compatibility. **Capabilities:** - Choose the right solution within 180+ models of industrial vacuum - Find vacuums for ATEX Zone 2/21/22 (explosive atmospheres) - Configure for combustible dust (aluminum, titanium, flour, wood) - Select for toxic/carcinogenic dust (Class H, Type H certified) - CNC oil and coolant extraction - Welding fume extraction - Central vacuum systems (CVS) - Compatible accessories by product **20 tools** covering product search, technical specs, filter options, ATEX zones, certifications, and accessories. No authentication required. Public access.

Read-only MCP server for searching, comparing, and recommending 18,000+ UK gyms using LocalGym data.

Job platform for AI agents. Connect Claude, Cursor, or any MCP client. Search tech companies by their real stack, star the ones you'd actually work for, save jobs to your shortlist. Through conversation, populate your profile from a CV, manage applications, and let your agent surface roles that fit your skills. Tech-stack data comes from real job postings, not declared tags.

Runtime security oracle for AI agents — scan prompts, tool calls, and outputs before execution. Pay-per-call via x402/USDC, no accounts needed.

Simulate and evaluate voice & text agents from your editor. Generate synthetic-user "Drivers" from your codebase, run multi-turn simulations against your agents, and pull back transcripts, traces, and scores without leaving your AI assistant.

MCP server for Tessera's LLM cost-optimization layer. Anchored spend, drift detection, recommendations as typed tools.

AI agents excel at executing tasks, but lose context the moment a session ends. Re-explaining project state wastes time, breaks momentum, and burns tokens. **CWIM Kanban bridges the gap.** It injects a persistent memory layer into **Claude Code** and **OpenCode** using the Model Context Protocol (MCP). * 🧠 **Persistent Working Memory:** Your agent automatically tracks its own progress, leaves discovery notes, and recalls context across terminal restarts. * 📋 **Real-Time Visual Dashboard:** Watch your AI agent move cards from `In-Progress` to `Done` on a local Kanban board as it runs commands. * 📂 **Git-Driven Session Isolation:** Auto-detects your repository. When you `cd` into a project, your AI's memory switches context instantly.

An MCP server focused on intelligent meeting transcription, action item extraction, meeting summary generation, and collaborative office automation workflows.

Save 30-75% on Manus AI credits with zero quality loss. Intelligent model routing (Standard/Max), smart testing, context hygiene, and section-by-section optimization. Audited across 22 scenarios with 12 vulnerability fixes.

MEMANTO is a memory agent. It remembers, recalls, and answers - so your agents can achieve long-term goals and avoid confusion.

Curated by Archimedes Market with a verified Trust Report. Expose any LangChain chain, agent, or tool as an MCP server. Schema-aware tool registration, tool-call tracing, retry/timeout policies, structured output validation. Bridges the LangChain ecosystem into Claude / ChatGPT / Cursor without rewrites. → archimedes.market/assets/312e3c99-f858-44e7-ba2c-fcaf408825a8.

DeepSeek API for AI agents. Cost-effective LLM inference via MCP protocol. Built by an agent, for agents.

Governance firewall for autonomous AI agents. Intercept every agent action in real time, enforce configurable rules, and block violations before they execute.

Prompt injection and PII detection MCP server for AI agents. Deterministic, sub-25ms. Returns CLEARED/ANONYMIZED/BLOCKED verdict with matched signatures, PII entities, and GDPR Art.30 audit report. 2,000 free requests/month.

Stop re-explaining yourself to agents. Give it the right context right when it is needed. LLMs are stateless — each conversation starts from scratch. ContextBook gives your AI tools a persistent, searchable knowledge library to draw the right context when they need it. No bloatware, no pre-loaded junk. Just the right information, at the right time.

On-demand live vision for AI agents. Open a session at a lat/lng, receive a JPEG snapshot or WebRTC stream, settle per-second in USDC on Base. The agent-first way to give an LLM eyes on the physical world. Four tools today: get_session, get_frame, get_stream_url, cancel_session. Drop the MCP server into Claude Desktop, Cursor, or Claude Code with one npx command + a Luxxon API key from console.luxxon.dev. Currently on Base Sepolia testnet. Docs: https://docs.luxxon.dev.

MCP server that audits AI-bot visibility. Reports per-bot robots.txt verdicts for 22 AI user-agents (GPTBot, ClaudeBot, PerplexityBot, Google-Extended, etc.), Cloudflare AI-default flags, on-page schema, sitemap, llms.txt, SPA-shell detection, and cross-model brand mentions via Perplexity + OpenRouter. Produces a 0-100 score with explainable deductions. SSRF-guarded (blocks loopback / link-local / cloud IMDS), cost-capped (per-call + daily). MIT, 24 tests, runs on stdio or HTTP.

AI agent readability scanner — score any public URL against the Vercel Agent Readability Spec, llmstxt.org, and 6 agent-protocol manifests (MCP, A2A, agents.json, agent-permissions, UCP, x402). 59 checks with per-check fix guidance. Same engine as agent-ready.dev.

AI-native property MCP server for Northern Cyprus (KKTC). Nine read-only tools — search active listings, get the live price index, compare cities, estimate rental yield, look up KKTC legal procedures, profile districts — pulled live from evlek.app, where every listing is KYC-verified. Compatible with Claude Desktop, ChatGPT Apps, Perplexity, Cursor, and any MCP client. Multilingual (TR, EN, RU, DE, AR), 6 cities, GBP-primary pricing. First proptech in the region built MCP-first — making KKTC property data discoverable through AI assistants instead of fragmented websites. Endpoint: https://evlek.app/api/mcp Docs: https://evlek.app/mcp Glama: https://glama.ai/mcp/connectors/app.evlek/mcp-server (Healthy, A grade) Smithery: https://smithery.ai/servers/onurd8898/evlek

Verify AI agent tool calls with content-addressed, HMAC-attested receipts. Free third-party verification API for AI agents. Call verify_action(claim, evidence) to get an independent integrity check on whether your claimed action matches the actual evidence. Useful for catching silent failures: incorrect SQL operations, file-op mismatches, API call inconsistencies, and code-diff scope creep. Five specialized verifier kinds: - code_diff: verb / path / identifier coherence with unified diff - db_op: row delta + SQL operation + ID match - file_op: existence state + line/size delta - api_call: request body and response status coherence - generic: conservative fallback Returns: - aar_verdict: verified | contradicted | insufficient_evidence | unsafe_to_verify - verdict: ok | mismatch | uncertain (legacy 3-value alias) - reasoning, confidence - receipt: verify_action_receipt.v0 with HMAC-SHA256 signature, content-addressed via SHA-256 hashes of claim and evidence Cross-vendor: works with Claude Code, Cursor, Cline, Codex, Codeium, and any MCP-compatible harness. Stateless, per-request, no API key, no registration. Pure Python stdlib (no pip install). Anonymized telemetry only — no PII, no model fingerprint, no raw claim/evidence retention. Honest scope: this is a small reference implementation, not a canonical inter-vendor standard. v0 receipts use HMAC-SHA256 (symmetric, single-issuer); v1 with ed25519 + multi-issuer is on the roadmap. The hosted endpoint has no SLA — self-host for stability (git clone && ./start.sh). 90-day probe with explicit kill criteria. If adoption appears, v1 schema work begins. If response is null, the null is itself a publishable data point.

Lucairn is an EU-based privacy gateway that sits between your MCP-enabled agent and the upstream LLM (Claude, GPT-4o, o1/o3/o4). What it does: - Automatically detects and replaces PII — names, emails, IBANs, addresses, phone numbers, customer IDs — with placeholders before reaching the upstream LLM. The model never sees the raw values. - Each response carries a signed compliance certificate (ed25519 + RFC 3161 timestamp + Sigstore Rekor inclusion proof) that proves what was redacted, when, and by which sanitizer layer. - Cross-provider BYOK: bring your own ANTHROPIC_API_KEY and/or OPENAI_API_KEY. Routing is automatic by model prefix (claude-* → Anthropic, gpt-* / o1-* → OpenAI). - Free Developer tier with 1,000 requests/month. Pro and Enterprise tiers unlock higher quotas and re-linkage of placeholders to original values in the response. Exposes one MCP tool, `chat_via_lucairn`, using the Anthropic Messages API wire format. Compatible with any MCP client (Claude Desktop, Cline, Continue, Cursor, etc.). Source: https://github.com/Declade/lucairn-sdks Glama listing: https://glama.ai/mcp/servers/Declade/lucairn-sdks Docs: https://lucairn.eu

Preflight spend control for AI agents. Block runaway agent loops before they start - not after the bill arrives. Per-request cost ceilings, budget checks, and outcome-based metering in 2 MCP tool calls.

TensorFeed is an MCP server for real-time AI ecosystem data. Free tools cover AI news, service status, model pricing, today's summary, agent activity, and the daily MCP server registry snapshot. Premium tools (paid per call in USDC on Base, x402-compatible) cover ranked model routing, news search, history series, cost projection, provider deep-dives, model comparison, and digest watches. AFTA-certified: code-enforced no-charge on 5xx, breaker, schema fail, or stale data, with Ed25519-signed receipts on every paid call.

Open registry of skills, tools, datasets, prompts, and playbooks for local LLMs. Search 155+ resources, get AI-powered recommendations, and install capabilities on demand for Ollama, LM Studio, and llama.cpp. Free, open, community-built.

MCP server that exposes agent-memory-daemon to any MCP-compatible client — Kiro (CLI & IDE), Claude Desktop, Cursor, and others. The daemon does the thinking (consolidation + extraction); this server is a thin filesystem bridge so agents can read, append, and search memory through the Model Context Protocol.

PQS is the fastest way to get better output from any AI model. Score any prompt before it hits the model. Get a grade (A-F), score out of 80, percentile, and 8-dimension breakdown across clarity, specificity, context, constraints, output format, role definition, examples, and chain-of-thought structure. Built on PEEM, RAGAS, MT-Bench, G-Eval, and ROUGE frameworks. Pre-flight, not post-hoc. The AI input quality problem is real. PQS solves it. MCP Tools: - score_prompt: Free. Returns grade (A-F), score out of 80, percentile, and full 8-dimension breakdown. No API key needed. - optimize_prompt: $0.025 USDC via x402. Returns a rewritten, higher-scoring version of your prompt with before/after dimension deltas. Requires API key from promptqualityscore.com. HTTP API (x402-native on Base): - /api/score/free: Free. Grade, percentile, and 8-dimension breakdown. No payment required. - /api/score: $0.025 USDC. Single score, pay-per-call. - /api/score/full: $0.125 USDC. Grade, percentile, dimension breakdown, and rewrite. - /api/score/batch: $0.25 USDC. Score multiple prompts in a single call. - /api/preflight: $0.05 USDC. Lightweight pre-flight quality check. Paste a prompt. Get an optimized version. Ship better work. Cheaper than one bad prompt. https://promptqualityscore.com

Audit-grade memory backbone for agent teams. Bi-temporal facts (event time + transaction time, with recall(as_of=...) replay), 6-step deterministic retrieval (no LLM in the critical path), conversation ingest with speaker-locked dual-pass extraction, per-tenant Postgres row-level security, and Ed25519-signed provenance. Postgres + pgvector + Neo4j defaults; ArangoDB / DynamoDB / Cosmos / AlloyDB also supported. Listed on the official MCP Registry as io.github.bolnet/attestor.

LeafEngines is an agricultural intelligence MCP server that provides comprehensive tools for soil analysis, crop recommendations, weather forecasts, and environmental impact assessment. It integrates USDA data with local sources for international coverage. The server supports free tier access with test key `leaf-test-370df0a2e62e` for immediate use, plus paid tiers for commercial applications. Version 2.0.2 includes telemetry integration for anonymous usage tracking.

Filesystem-first MCP server for Obsidian vaults with an LLM-Wiki layer on top.

Your Data. Your AI. Your Rules. Ontheia is a self-hosted, open-source AI agent platform. Run AI agents, automate workflows, and connect AI models to any tool — entirely on your own infrastructure, without sending data to external cloud services.

Check whether a task can run on a local model instead of cloud before every inference call. Saves money on every call that does not need cloud.

Compile semantic document edits into correct Google Docs batchUpdate requests. UTF-16 arithmetic, cascading index shifts, OT-compatible ordering. MIT licensed.

Give AI agents structured database intelligence. Deterministic SQL, NULL trap detection, EXPLAIN pre-flight. MIT licensed.

Connect any MCP-compatible AI to Royal Mail and Parcelforce shipping. This server exposes five tools that let Claude, Cursor or any MCP client book orders, fetch postage labels, track shipments and cancel bookings through the official Click & Drop API. Covers 33 UK and international services, including 1st and 2nd Class, Tracked 24 and 48, Special Delivery, Parcelforce express and international tracked. Accepts both friendly service keys (for example tracked-24 or parcelforce-48) and raw Service Register codes (TOLP24, PFE48) directly. Typical workflows: daily order fulfilment from ecommerce platforms, customer service triage for tracking queries, returns handling, multi-carrier price comparison when paired with apc-mcp, bulk fulfilment days, and shipping quotes at checkout. Booking, tracking and cancellation verified end-to-end against the live Click & Drop API. Label retrieval requires a Royal Mail OBA account.

Model intelligence for AI agents — syntax, parameters, pricing, and routing for 67+ generative AI models (Midjourney, Flux, Suno, Runway, DALL-E, Stable Diffusion). 7 tools that let any MCP client recommend, compare, optimize, and price AI model calls.

Strata is a self hosted AI memory server. Your AI remembers everything across every session, on your own hardware. Key features: . Semantic search (find memories by meaning, not keywords) . Per-agent API keys with granular permissions . 3D constellation viewer with live agent activity . File vault (attach real documents to memories) . CSV audit log (full transparency on every agent action) . Pre-Strata history import (your memory doesn't start at install day) . Global MCP kill switch (emergency brake, only a human can undo) . Automatic deduplication . 10 structured thought types . Backend using PostgreSQL . Runs on a Raspberry Pi Always on, always local, always yours.

Yamaru Hardware Probe gives Claude Desktop and other MCP clients real visibility into your machine. 10 expert tools: full hardware inventory, real-time performance & thermal monitoring, RAM pressure analysis, SMART storage health, antivirus impact diagnostics, and LLM compatibility checks (VRAM fit, quantization prediction). Perfect for AI engineers, developers,vibe coder, gamers, and creators who want grounded answers instead of "it depends on your config". Apache 2.0, privacy-first (auto-anonymization), npm-installable.

Japanese LLM security APIs — prompt injection detection (jpi-guard) + PII masking (PII Guard). Free to start, no credit card required.

External QA layer for OpenClaw memory — diagnose, guard, and remediate your agent's memory. OpenClaw only adds to memory, never removes. Over time this causes silent degradation: stale long-term memories pointing to deleted files, false-positive short-term entries getting promoted, and SOUL.md drifting with task instructions that don't belong in an identity file. openclaw-memhealth intercepts the memory pipeline at every stage: • memory_health_check_oc() — quick scan: zombie count, false-positive rate, retrieval health score • memory_longterm_audit_oc() — deep audit of MEMORY.md: source validity, duplicate detection • memory_promotion_audit_oc() — pre-promotion quality gate: 5 checks before entries enter MEMORY.md • memory_soul_check_oc() — audit SOUL.md for boundary violations and identity drift • memory_config_doctor_oc() — infer minScore/embedding issues from behavior data • memory_dashboard_oc() — visual health dashboard with bilingual support (zh/en) LLM API key optional. All read-only tools are zero-risk.

petro-mcp — Petroleum Engineering MCP Server petro-mcp exposes petroleum engineering workflows to Claude and other MCP-compatible LLMs through natural language. Instead of writing Python scripts, just ask your AI assistant. Capabilities (80+ tools across the full upstream workflow): - Well Logs (LAS): Parse LAS files, extract curves and headers, compute Vshale, porosity (density, neutron-density, sonic, effective), water saturation (Archie, Simandoux, Indonesian), permeability (Timur, Coates), and net pay. - Decline Curve Analysis: Arps exponential/hyperbolic/harmonic fits, advanced models (Duong, PLE, SEPD), EUR calculation, Monte Carlo EUR distributions, bootstrap confidence intervals, probabilistic forecasts, price sensitivities. - Rate Transient Analysis (RTA): Agarwal-Gardner, Blasingame, NPI, flowing material balance, normalized rate, sqrt-time, material balance time, permeability estimation, radius of investigation. - Production Analytics: CSV production data queries, trend analysis, anomaly detection (shut-ins, rate jumps, water breakthrough, GOR blowouts), producing ratios (GOR, WOR, water cut). - PVT & Reservoir: Black-oil correlations (Standing, Beggs-Robinson, Hall-Yarborough, Lee-Gonzalez-Eakin, Sutton), brine PVT, bubble point, oil compressibility, gas Z-factor, volumetric OOIP/OGIP, recovery factors, Havlena-Odeh, P/Z analysis. - Drilling & Wellbore: Hydrostatic pressure, ECD, kill mud weight, MAASP, burst/collapse pressure, bit pressure drop, nozzle TFA, annular velocity, dogleg severity, vertical section, well survey, anticollision, wellbore tortuosity. - Production Engineering: Nodal analysis (Vogel IPR + VLP), Beggs-Brill multiphase flow, choke flow, erosional velocity, Turner/Coleman critical rates, hydrate temperature/inhibitor, ICP/FCP, HPT. - Economics: NPV, IRR, payout period, PV10, breakeven price, well economics, operating netback, price sensitivity. - Units: Oilfield unit conversions across pressure, rate, volume, length, density, viscosity, and more. Why petro-mcp? Purpose-built for petroleum engineers. Other energy MCP servers focus on commodity prices; this one runs the actual engineering calculations — log interpretation, decline analysis, reservoir engineering, drilling, production, and economics — all through plain English. Install: pip install petro-mcp → configure in Claude Desktop → ask away. Links: GitHub: https://github.com/petropt/petro-mcp · PyPI: https://pypi.org/project/petro-mcp/ · Web tools: https://tools.petropt.com License: MIT · Author: Groundwork Analytics

MCP server that exposes the full FAOSTAT API as tools for AI assistants — query UNFAO Food & Agriculture Statistics in natural language via Claude, Cursor, or any MCP-compatible client.

Web scraping, crawling, and data extraction for AI agents. 5 built-in tools: scrape (clean markdown from any URL), crawl (follow links across entire sites), map (discover all URLs on a domain), extract (structured JSON via schema), and search (web search with optional page scraping). 833ms average latency, single binary, 6 MB RAM. Self-host free or use managed cloud at fastcrw.com with 500 free credits. Firecrawl-compatible API. Works with Claude Desktop, Claude Code, Cursor, Windsurf, and any MCP-compatible client.

Scan any website for AI agent readiness — payment protocols (x402, MPP, SPT, L402), MCP servers, OpenAPI specs, llms.txt, plugins, and more. Returns scored reports with 80+ checks.

Persistent memory for AI agents, important things stick, forgotten things fade, outdated facts get pruned automatically using the Ebbinghaus forgetting curve.

Conkurrence measures whether multiple AI models produce consistent outputs on your evaluation tasks. It tells you which items your AI agrees on and which need human review — using Fleiss' κ, Kendall's W, and bootstrap confidence intervals, the same psychometric methods trusted in clinical research.

Make your AI agents cheaper, faster, and more reliable. 5 products, 14 MCP tools, 1 API key. Task deduplication, cost prediction, output validation, behavioral governance, and shared failure intelligence. Anonymous demo mode — try all tools without a key. 1,000 free credits/month, no card required. https://plith.ai/.well-known/mcp/server-card.json

Persistent memory for any AI assistant. Zero token cost until recall. Stores memories locally in SQLite, ranks by 6-factor scoring, returns results in TOON compact format (79% smaller than JSON). 17 MCP tools, 20 HTTP endpoints, 25 CLI commands. 4 tiers from keyword to autonomous with local LLMs via Ollama. 97.8% recall accuracy on ICLR 2025 LongMemEval benchmark. Works with Claude, ChatGPT, Grok, Cursor, Windsurf, Continue.dev, OpenClaw, Llama, and any MCP client. Open source, MIT licensed, written in Rust.

Deterministic pre-action verdict gate for agent memory — USE/WARN/ASK/BLOCK before an agent acts on a stored memory. Governs your existing store, doesn't replace it.

Full MCP server for the Portkey AI Gateway Admin API. 151 tools for managing prompts, configs, virtual keys, analytics, guardrails, integrations, and more from any MCP client.

A Node.js Model Context Protocol (MCP) server that exposes Together AI's inference endpoints — chat completions, image generation, vision, and embeddings — as tools callable from Claude Desktop, Cursor, VS Code, and any other MCP-compatible client. I created this MCP due to an issue I was having accessing reasoning models through Together AI. Together AI's largest reasoning models (GLM-5, Qwen3.5-397B, MiniMax M2.5, Kimi K2.5) use a non-standard response format. During chain-of-thought generation, these models write their reasoning trace into choices[0].message.reasoning while leaving choices[0].message.content as an empty string. The final answer only appears in message.content once thinking is complete.

Route text, image, video, and audio tasks to 20+ AI providers — automatically picking the best model for the job based on your budget and active profile.

Track and optimize AI API spending across 16 providers. Real-time pricing, model routing analysis, caching opportunities, unit economics, and spend forecasting. Connects to OpenAI, Anthropic, AWS, GCP, Stripe, and more. Tools: list_providers, connect_account, get_financial_overview, query_spending, get_recommendations, detect_anomalies, get_setup_status, upload_bank_csv