gitrakz

gitrakz is a self-hosted GitHub activity tracker. It syncs a user's gh activity into a local SQLite database, renders a filterable timeline and derived work sessions, and runs programmatic templates — deterministic transform pipelines over the timeline that export to CSV, PDF, or JSON. One Go binary with the Svelte UI embedded; no external services required.

For setup — the installer, the direct-Docker path, config, and auth — read references/setup.md before touching a stack.

Security and safety

• This skill drives a gitrakz instance the user already runs. Take GITRAKZ_URL (e.g. http://127.0.0.1:8080) and, if set, GITRAKZ_AUTH_TOKEN from the environment or ask — do not hunt the workspace for tokens.
• gitrakz authenticates to GitHub with a token from gh auth token, injected at runtime and never written to disk. It reads activity only; mount a read-scoped token where you control the scope.
• POST /api/v1/sync starts a GitHub sync (network + rate-limit spend) and POST /api/v1/run may call an LLM if the template uses describe-work or a prose block. Only trigger these for the task the user named.
• GITRAKZ_ELELEM_BASE_URL receives commit titles / diffs when LLM steps run. Point it only at endpoints the user trusts. Template output is typed display blocks, never author-supplied HTML.
• gitrakz also exposes its capabilities as MCP tools (see "MCP" below), over the same two surfaces: streamable HTTP at /mcp (Bearer-gated exactly like /api/v1 when GITRAKZ_AUTH_TOKEN is set) and stdio via the binary's mcp subcommand. Every MCP tool wraps the same read-mostly service layer as the REST API above — same sync/LLM cost caveats apply to gitrakz_trigger_sync and gitrakz_run_template.

Use it for

• Installing or running gitrakz and confirming it is up.
• Triggering an incremental sync (POST /api/v1/sync) and checking progress (GET /api/v1/sync/status).
• Pulling a filtered timeline or a derived work-sessions timesheet.
• Listing, creating, editing, and running templates, and exporting a run to CSV / PDF / JSON.
• Any of the above through MCP tools instead of raw REST calls, when the driving client speaks MCP (see "MCP" below).

Do not use it for

• Writing to GitHub — gitrakz only reads activity.
• Any instance the user does not run and trust.
• Treating /api/v1/run output as HTML — it is a typed block document.

Talk to a running instance

All endpoints are under /api/v1, JSON, camelCase. When GITRAKZ_AUTH_TOKEN is set on the server, send Authorization: Bearer <token>; when it is empty the API is open (single-user / trusted network). The SPA at / needs no token.

: "${GITRAKZ_URL:=http://127.0.0.1:8080}"
auth=(); [ -n "${GITRAKZ_AUTH_TOKEN:-}" ] && auth=(-H "Authorization: Bearer $GITRAKZ_AUTH_TOKEN")

# Is it up? (SPA, unauthenticated)
curl -fsS "$GITRAKZ_URL/" >/dev/null && echo "gitrakz is up"

# Trigger a sync, then poll status.
curl -fsS "${auth[@]}" -X POST "$GITRAKZ_URL/api/v1/sync"
curl -fsS "${auth[@]}" "$GITRAKZ_URL/api/v1/sync/status"

# Distinct owners, then repos under one.
curl -fsS "${auth[@]}" "$GITRAKZ_URL/api/v1/owners"
curl -fsS "${auth[@]}" "$GITRAKZ_URL/api/v1/repos?owner=OWNER"

# Timeline (paginated with hasMore — never a total) and derived sessions.
curl -fsS "${auth[@]}" "$GITRAKZ_URL/api/v1/timeline?owner=OWNER&from=2025-01-01&perPage=50"
curl -fsS "${auth[@]}" "$GITRAKZ_URL/api/v1/sessions?owner=OWNER"

Endpoints

Everything the SPA does is one of these calls:

• GET /api/v1/owners — distinct owners.
• GET /api/v1/repos?owner= — repos under an owner.
• GET /api/v1/timeline?owner=&repo=&type=&from=&to=&page=&perPage= — events; type is one of commit|pr|review|issue|release. Paginated with hasMore.
• GET /api/v1/sessions?… — sessionized view + heuristic hours.
• POST /api/v1/sync — trigger an incremental sync.
• GET /api/v1/sync/status — last sync status.
• GET /api/v1/templates — list (built-in + custom).
• POST /api/v1/templates — create a custom template.
• PUT /api/v1/templates/{id} — edit (built-ins clone-on-edit).
• DELETE /api/v1/templates/{id} — delete a custom template.
• POST /api/v1/templates/generate — LLM-compose a template draft from a prompt.
• POST /api/v1/run — run a template over a filter → a typed block document.
• POST /api/v1/export — export a document / run to csv | pdf | json.

The OpenAPI spec at api/api.yml in the repo is the source of truth for request and response shapes.

MCP

gitrakz exposes the same capabilities above as MCP (Model Context Protocol) tools — no REST calls needed if the client speaks MCP. Both transports serve identical tools against the same running instance's SQLite data.

Tools

• gitrakz_list_owners — every owner with ingested activity. No input.
• gitrakz_list_repos — repos under owner (required).
• gitrakz_list_templates — every saved template (built-in + custom). No input.
• gitrakz_get_template — one template by id (required).
• gitrakz_run_template — run templateId (required) over an optional filter (owner/repo/type/from/to) and formValues; returns the rendered document as typed display blocks (never HTML).
• gitrakz_trigger_sync — trigger one incremental gh sync. No input. Same network + rate-limit cost as POST /api/v1/sync.
• gitrakz_get_sync_status — current sync status. No input.
• gitrakz_list_sessions — derived work sessions over an optional owner/from/to filter.
• gitrakz_query_timeline — one page of the filtered, newest-first event timeline (owner/repo/type/from/to/page/perPage).

Streamable HTTP — same host as the REST API

Mounted at /mcp on the running instance, alongside /api/v1. Bearer-gated the same way: send Authorization: Bearer <token> when GITRAKZ_AUTH_TOKEN is set on the server.

{
  "mcpServers": {
    "gitrakz": {
      "type": "http",
      "url": "http://127.0.0.1:8080/mcp",
      "headers": { "Authorization": "Bearer <GITRAKZ_AUTH_TOKEN>" }
    }
  }
}

Omit headers when the server has no GITRAKZ_AUTH_TOKEN set.

stdio — local Claude Code use

The Go binary's mcp subcommand (not the shell wrapper's gitrakz start / stop / etc. commands) opens the exact same SQLite database and runs the MCP server over stdio. Since gitrakz ships as a container, run it through Docker with -i for a live stdin/stdout pipe, against the same named data volume the stack already uses:

{
  "mcpServers": {
    "gitrakz": {
      "command": "docker",
      "args": [
        "run", "--rm", "-i",
        "-e", "GH_TOKEN",
        "-v", "gitrakz-data:/data",
        "psyb0t/gitrakz:vX.Y.Z", "mcp"
      ],
      "env": { "GH_TOKEN": "<gh auth token>" }
    }
  }
}

Pin the same released tag the running stack uses. GH_TOKEN is only needed for gitrakz_trigger_sync; every read-only tool works without it.

Setup

Everything about installing, running with Docker directly, configuring ~/.config/gitrakz/.env, and the GH_TOKEN auth model lives in references/setup.md.