Docs
Supported tools
One ledger, cost, and grade across every coding agent your team runs - not just Claude Code.
Adapters shipped today
RepoOps registers a session-capture adapter per tool. Every adapter feeds the SAME pipeline (the Activity Ledger, cost normalization, the accountability loop) - no per-tool silo.
- Claude Code - first-class; reads
CLAUDE.mdnatively with zero setup. - Cursor - format-verified against a real install.
- Codex - experimental (awaiting a live install to verify against).
- GitHub Copilot, Continue, Xcode - parsers shipped.
- Cline, Aider, Roo Code, Windsurf - editor-store adapters.
- Claude Desktop, Claude Cowork, Claude Routines - additional Claude surfaces.
- Generic ingest -
POST /v1/traces(OpenTelemetry) orPOST /api/sessions/ingest(plain JSON) for anything else.
Bring your own agent: the external-adapter protocol
Not on the list above? Write a small standalone program that speaks one protocol, and RepoOps captures its sessions without a code change on either side.
Declare it in .repoops-external-adapters.json at your repo root:
[{ "name": "my-agent", "command": "/path/to/binary", "args": [] }]RepoOps spawns the command, writes {"repoRoot": "...", "sinceMs": <n|null>} to its stdin, and reads a JSON array of session objects back from stdout. A reference implementation lives at scripts/external-adapters/reference-adapter.mjs.
Where the cost shows up
The Activity Ledger's per-source cost rollup (a chip per tool, scaled to the highest spender) reads every adapter above with no extra setup.