OpenLIT CLI
OpenTelemetry-based CLI tracing for Claude Code, Cursor and Codex sessions, showing what the agent actually did and what it cost
- 3,057 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- Collects and stores traces of agent calls, including request content
- Requires a self-hosted backend with a database
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/openlitDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Collects and stores traces of agent calls, including request content
- Requires a self-hosted backend with a database
This is third-party code. Review the repository files before installing.
What it does
OpenLIT runs locally via docker compose (including ClickHouse for trace storage) and shows a dashboard at localhost:3000. A separate CLI, installed via install.sh, enables coding-agent tracing with openlit coding install --vendor=all for Claude Code, Cursor and Codex. Traces are sent to a configured OTLP endpoint and show LLM calls, tool calls, prompts, cost and errors for a session. Separate Python and TypeScript SDKs (openlit.init()) instrument your own AI applications outside of coding agents.
Who it is for. For developers and teams running coding agents on real projects who want visibility into what the agent did and what it cost.
Good fit when
- You need to trace tool calls and LLM calls in a coding-agent session
- You want to track cost and errors across several agents at once
- You need to evaluate LLM, tool and retrieval steps in your own AI application
Not a fit when
- You do not want to run your own docker compose stack with ClickHouse just for basic stats
- The logs already provided by the agent or its vendor are enough
Example request
Show me the trace of the last Claude Code session: which tools were called and what it costLimitations
The dashboard backend is a self-hosted docker compose stack that you need to host yourself. CLI instrumentation covers specific coding-agent vendors; --vendor=all wires up whatever is supported at install time. Both the CLI and the SDK need to be pointed at the same OTLP endpoint separately.
How to disable. Stop the stack: docker compose down, remove the CLI and coding-agent tracing configuration, and remove openlit.init() calls from your code.
FAQ
Do you need a separate server?
Yes, the dashboard and trace storage run locally via docker compose, including ClickHouse.
Does this work for regular AI applications, not just coding agents?
Yes, separate Python and TypeScript SDKs instrument LLM providers and frameworks in any AI application through openlit.init().
Discussion
No questions or reviews yet.
Related
CLI toolsA local dashboard for Claude Code, Codex CLI, Gemini CLI, Copilot and OpenCode that shows tokens, cost, latency and errors
An htop-like terminal panel for AI agents: Claude Code, Codex, OpenCode and Devin sessions, CPU, tokens and cost.
A terminal dashboard for Claude Code and Codex: a tree of agents with live status, a tool-call feed and context headroom.
A local telemetry collector for AI coding agents: tokens, sessions and tool calls in one OpenTelemetry-style event schema
A self-hosted control center for coding agents: Claude Code, Codex, Copilot, OpenCode and Pi from your phone, desktop and terminal
The LangSmith plugin sends Claude Code conversations to LangSmith so you can see model calls, tools and subagents
README in short
The README describes OpenLIT as an open-source observability and evaluation platform for AI applications and coding agents built on OpenTelemetry. A five-minute quickstart brings up the stack via docker compose plus an SDK for regular applications, and a separate section explains installing a CLI that specifically instruments Claude Code, Cursor and Codex sessions and sends traces to the same endpoint. The project also ships as a plugin marketplace entry for Claude Code, Codex and Cursor.