opentraces CLI
opentraces records Claude Code, Codex and Pi agent work into a local log, anchors edits to commits and prepares datasets.
- 104 ★ on the repository
- Last commit
Extended access
Why this level
- Session records may contain secrets and customer data, and publishing requires an explicit choice.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/opentracesDetects 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.
- Session records may contain secrets and customer data, and publishing requires an explicit choice.
This is third-party code. Review the repository files before installing.
What it does
opentraces is a local evidence layer for agent work: an open schema and a CLI. Through session hooks it captures what the agent saw, did and changed, and stores the record in a private local bucket: trajectories, patch history, context events and blobs. Three linked records: Trace answers what the agent did, Trail shows which edits reached a commit and survived (`trail blame`, `trail graph`), and Ctx reconstructs what the model saw at each step. The log supports search, slices, context-waste signals, run comparison and session resumption. Separate workflows turn trajectories into dataset rows, and security checks run before publishing to the Hugging Face Hub.
Who it is for. Developers who accumulate agent work traces for debugging, training and incident analysis and want to keep that data in their own hands.
Good fit when
- Which agent edits survived review and reached a commit must be found.
- A session must be resumed with the context the model saw at a given step.
- A dataset for evaluation or training is built from working sessions.
- Two agent runs must be compared by context waste.
Not a fit when
- Agent sessions contain secrets and sharing them anywhere is forbidden.
- Simple telemetry without storing session content is wanted.
- The agent used is not Claude Code, Codex or Pi.
Example request
Using opentraces, show which patches from my last session reached a commit and print a context waste summary.Limitations
The log stores whole trajectories, so it may contain secrets, credentials, internal paths and customer data; the authors warn that publishing reduces but does not remove this exposure and advise reading the security tiers docs before uploading. Hugging Face sync is enabled only with explicit consent. The schema and CLI are young, the project license shows as NOASSERTION on GitHub, and three agents are supported at the moment. `opentraces uninstall` removes every installed surface.
How to disable. Run `opentraces uninstall`: it reverses hooks and extensions for Claude Code, Codex and Pi. Then remove the package (`pipx uninstall opentraces`) and the bucket directory if the records are no longer needed.
FAQ
How do I install it?
Run `pipx install opentraces`. `brew install JayFarei/opentraces/opentraces` and `npx skills add jayfarei/opentraces` for the skill also exist.
Which agents are supported?
Claude Code, Codex CLI and Pi, through session hooks and extensions.
What are Trace, Trail and Ctx?
Trace describes what the agent did. Trail links patches to commits. Ctx shows what the model saw at each step.
Does data leave my machine?
No, until you explicitly enable sync to a Hugging Face remote.
How do I remove it?
With `opentraces uninstall`, then remove the package the way you installed it.
Discussion
No questions or reviews yet.
Related
CLI toolsA local trajectory compiler that turns Claude Code and Codex session history into training data, tasks and verifiers
A local whiteboard and MCP server for research agents: every claim is tied to the run and the code that produced it
A skill and web viewer that turn Claude Code, Codex and Pi logs into a decision map of the agent run with links to originals.
A CLI that reads local Claude Code, Codex and Hermes sessions, finds repeated failures and stages fixes for manual review
A shared board for Claude Code, Codex and other agents: a question an agent stops on rises to the top within seconds.
A multi-agent harness that verifies results by artifacts: stop hooks, an independent judge and an event log for Claude Code, Codex, Cursor and more
README in short
A log of agent sessions that ties edits to commits and replays context. Data stays local until sync is enabled.