Ingero

v0.8.2Monitoring & Observabilitystable

eBPF-based GPU causal observability agent with MCP server. Traces CUDA Runtime and Driver APIs via kernel uprobes and host events via tracepoints to build causal chains explaining GPU latency. 7 tools: get_check, get_trace_stats, get_causal_chains, g

causal-tracingcudacuda-graphsebpfgpu
Share:
82
Stars
0
Downloads
0
Weekly
0/5

What is Ingero?

Ingero is a Model Context Protocol (MCP) server that allows AI assistants like Claude, Cursor, and VS Code to ebpf-based gpu causal observability agent with mcp server. traces cuda runtime and driver apis via kernel uprobes and host events via tracepoints to build causal chains explaining gpu latency. 7 tools...

eBPF-based GPU causal observability agent with MCP server. Traces CUDA Runtime and Driver APIs via kernel uprobes and host events via tracepoints to build causal chains explaining GPU latency. 7 tools: get_check, get_trace_stats, get_causal_chains, g

This server falls under the Monitoring & Observability category on MCPgee, the world's largest MCP server directory with 33,000+ servers.

Features

  • eBPF-based GPU causal observability agent with MCP server. T

Use Cases

GPU performance observability
eBPF-based causal tracing
CUDA latency analysis and incident response
ingero-io

Maintainer

LicenseApache-2.0
Languagec
Versionv0.8.2
UpdatedMay 18, 2026
Statushealthy
Maintenanceactive

Works with

ClaudeOpenAIwindowsmacoslinux

Installation

Manual Installation

npx ingero

Configuration

Configuration Details

Config File

claude_desktop_config.json

Performance

Response Metrics

Response Time< 200ms
ThroughputMedium

Resource Usage

Memory UsageLow
CPU UsageLow

How to Set Up and Use Ingero

Ingero is an eBPF-based GPU causal observability agent with an integrated MCP server that traces CUDA Runtime and Driver API calls via kernel uprobes and host events via tracepoints to build causal chains explaining GPU latency. It exposes seven MCP tools — including get_check, get_trace_stats, get_causal_chains, and others — enabling AI assistants to query live GPU performance data and diagnose latency bottlenecks in CUDA workloads running on Kubernetes or bare-metal hosts. This is designed for ML engineers, platform engineers, and SRE teams investigating GPU performance issues and incident response.

Prerequisites

  • Linux host with eBPF support (kernel 5.8+ recommended) and CUDA-capable GPU
  • CUDA Runtime installed on the target host where GPU workloads run
  • Root or CAP_BPF + CAP_PERFMON privileges to load eBPF programs
  • Kubernetes cluster (optional) if tracing containerized GPU workloads
  • An MCP-compatible AI client such as Claude Desktop or Claude Code
1

Review system requirements

Ingero uses eBPF uprobes to trace CUDA APIs. Verify your Linux kernel version and that the CUDA runtime is installed on the target machine.

uname -r
nvcc --version
2

Clone the Ingero repository

Clone the repository from GitHub to obtain the source code and deployment manifests.

git clone https://github.com/ingero-io/ingero.git
cd ingero
3

Deploy the Ingero agent

Follow the deployment instructions in the repository. For Kubernetes, apply the provided manifests. The agent requires privileged access to load eBPF programs.

# For Kubernetes:
kubectl apply -f deploy/
4

Start the MCP server

Once the agent is running and tracing CUDA calls, start the MCP server component to expose the seven observability tools to your AI client.

5

Configure your AI client to connect

Add the Ingero MCP server to your AI client configuration to enable GPU observability queries.

Ingero Examples

Client configuration

Connect an MCP client to the Ingero agent. Adjust the command and host details based on your deployment.

{
  "mcpServers": {
    "ingero": {
      "command": "npx",
      "args": ["ingero"]
    }
  }
}

Prompts to try

Use these prompts to investigate GPU performance issues via the Ingero MCP tools.

- "Check the current GPU health status using the ingero MCP."
- "Get trace statistics for the last 10 minutes of GPU activity."
- "Show me the causal chains explaining the GPU latency spike at 14:32."
- "Identify which CUDA kernel calls are contributing most to inference latency."
- "Is there a CUDA graph bottleneck in the current workload?"

Troubleshooting Ingero

eBPF program fails to load with permission denied

The Ingero agent requires elevated privileges to load eBPF programs. Ensure the process has CAP_BPF and CAP_PERFMON capabilities, or run as root. In Kubernetes, the DaemonSet must run with privileged: true in its security context.

No trace data is collected despite the agent running

Verify that CUDA workloads are actually running on the target host. The agent traces CUDA Runtime and Driver API calls via uprobes — if no CUDA processes are active, trace stats will be empty. Confirm the CUDA library path is correctly detected by the agent.

MCP server is unreachable from the AI client

Confirm the Ingero MCP server process is running and listening on the expected port. Check firewall rules if connecting across hosts. For Kubernetes deployments, ensure the MCP service is exposed via a NodePort or LoadBalancer service.

Frequently Asked Questions about Ingero

What is Ingero?

Ingero is a Model Context Protocol (MCP) server that ebpf-based gpu causal observability agent with mcp server. traces cuda runtime and driver apis via kernel uprobes and host events via tracepoints to build causal chains explaining gpu latency. 7 tools: get_check, get_trace_stats, get_causal_chains, g It connects AI assistants to external tools and data sources through a standardized interface.

How do I install Ingero?

Follow the installation instructions on the Ingero GitHub repository. Clone the repo, install dependencies, and add the server config to your AI client.

Which AI clients work with Ingero?

Ingero works with all major MCP-compatible AI clients including Claude Desktop, Claude Code, Cursor, VS Code (GitHub Copilot), Windsurf, and Cline.

Is Ingero free to use?

Yes, Ingero is open source and available under the Apache-2.0 license. You can use it freely in both personal and commercial projects.

Browse More Monitoring & Observability MCP Servers

Explore all monitoring & observability servers available in the MCPgee directory. Each server includes setup guides for Claude, Cursor, and VS Code.

Quick Config Preview

{ "mcpServers": { "ingero": { "command": "npx", "args": ["-y", "ingero"] } } }

Add this to your claude_desktop_config.json or .cursor/mcp.json

Read the full setup guide →

Ready to use Ingero?

Browse our complete directory of 33,000+ MCP servers, read setup guides for your editor, and start building with the Model Context Protocol.

33,000+ ServersFree & Open SourceStep-by-Step Guides