Ingero

Created By

ingero-io4 months ago

eBPF-based GPU causal observability agent with MCP server. Traces CUDA Runtime/Driver APIs and host kernel events to build causal chains explaining GPU latency

# ebpf

# gpu

Overview Content Tools Comments

Overview

Ingero is an eBPF-based agent that provides production-safe, kernel-level causal tracing for GPU workloads. It answers: "Why is my GPU training/inference slow right now?"

MCP Tools (7)

get_check — System diagnostics (kernel, BTF, NVIDIA, CUDA, GPU processes)
get_trace_stats — Per-operation p50/p95/p99 latency stats
get_causal_chains — Root cause analysis with severity ranking and fix recommendations
get_stacks — Resolved call stacks (Python source file, function, line number)
run_sql — Read-only SQL queries against the event database
run_demo — Run synthetic demo scenarios (no GPU or root needed)
get_test_report — GPU integration test results

How It Works

Ingero traces 4 layers via eBPF:

CUDA Runtime API (libcudart.so) — cudaMalloc, cudaFree, cudaLaunchKernel, cudaStreamSync
CUDA Driver API (libcuda.so) — cuLaunchKernel, cuMemcpy, cuCtxSynchronize, cuMemAlloc
Host kernel tracepoints — sched_switch, mm_page_alloc, oom_kill, process lifecycle
System context from /proc — CPU, memory, load, swap

These produce causal chains: System context + host event → CUDA call → root cause.

Usage

# Start MCP server (stdio — for Claude Code, Cursor, etc.)
ingero mcp --db ~/.ingero/ingero.db

# Start MCP server (HTTPS — for remote clients)
ingero mcp --db ~/.ingero/ingero.db --http :8090

Key Features

- <2% overhead, zero code changes, single binary
- Python 3.10/3.11/3.12 source line attribution via DWARF
- SQLite storage with 10 GB rolling cap
- Kubernetes support (DaemonSet, Helm chart, pod metadata)
- Tested on A10, A100, H100, GH200, RTX 3090, RTX 4090

Try in Playground

Server Config

{
  "mcpServers": {
    "ingero": {
      "command": "ingero",
      "args": [
        "mcp"
      ],
      "env": {}
    }
  }
}

Project Info

Created At

4 months ago

Updated At

4 months ago

Author Name

ingero-io

Star

Language

License

Recommend Servers

View All

Serper MCP Server

@garymengcom

A Serper MCP Server

Python

a year ago

EdgeOne Pages MCP

@TencentEdgeOne

An MCP service designed for deploying HTML content to EdgeOne Pages and obtaining an accessible public URL.

TypeScript

a year ago

3 days ago

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

JavaScript

a year ago

Guatemala Payments Mcp

@junter1989k-ai

3 days ago

Filesystem

@modelcontextprotocol

3 months ago

Puppeteer

@modelcontextprotocol

Browser automation and web scraping

a year ago

3 days ago

3 days ago

Framelink Figma MCP Server

@GLips

MCP server to provide Figma layout information to AI coding agents like Cursor

TypeScript

a year ago

Hong Kong Payments Mcp

@junter1989k-ai

3 days ago

Australia Payments Mcp

3 days ago

3 days ago

Playwright MCP server

TypeScript

a year ago

3 days ago

3 days ago

BlenderMCP connects Blender to Claude AI through the Model Context Protocol (MCP), allowing Claude to directly interact with and control Blender. This integration enables prompt assisted 3D modeling, scene creation, and manipulation.

a year ago

GBOX Android MCP

@babelcloud

GBOX provides environments for AI Agents to operate computer and mobile devices. Mobile Scenario: Your agents can use GBOX to develop/test android apps, or run apps on the Android to complete various tasks(mobile automation). Desktop Scenario: Your agents can use GBOX to operate desktop apps such as browser, terminal, VSCode, etc(desktop automation). MCP: You can also plug GBOX MCP to any Agent you like, such as Cursor, Claude Code. These agents will instantly get the ability to operate computer and mobile devices.

10 months ago

Memory

@modelcontextprotocol

a year ago

3 days ago

ContextBridge (CB) is a local-first retrieval layer that sits between your codebase and your AI coding agent. Instead of pasting raw source files into context or letting the AI guess which files matter, CB indexes your codebase's real structure (via Graphify) and returns a compact, ranked result — owner file, related files, key symbols, and dependency chains — grounded in your actual code. A typical response is ~4–8 KB instead of the 100+ KB of raw source an AI would otherwise need to read to answer the same question — roughly a 96% cut in input tokens sent to your cloud AI, with zero hallucinated file paths or method names.

3 days ago

Ukraine Payments Mcp

@junter1989k-ai

3 days ago

Brave Search

@modelcontextprotocol

Web and local search using Brave's Search API

a year ago

Costa Rica Payments Mcp

@junter1989k-ai

3 days ago

Kazakhstan Payments Mcp

3 days ago

3 days ago

A Model Context Protocol (MCP) server for image generation and manipulation using fal.ai's Stable Diffusion model.

a year ago

Perplexity Ask MCP Server

@ppl-ai

A Model Context Protocol Server connector for Perplexity API, to enable web search without leaving the MCP ecosystem.

JavaScript

a year ago

3 days ago

Ready for a new way to bill pay? Pay any bill in a snap, right from a chat. Describe or snap a photo of your bill (electric, water, gas, internet, phone, medical, credit card, rent, parking tickets, traffic violations and more) and Papaya reads it, then hands you a secure link to pay by card. Fast, secure, and no juggling twelve logins, with full or partial payments and status updates. Powered by Papaya (papayapay.com).

3 days ago