JSPM

  • ESM via JSPM
  • ES Module Entrypoint
  • Export Map
  • Keywords
  • License
  • Repository URL
  • TypeScript Types
  • README
  • Created
  • Published
  • Downloads 2378
  • Score
    100M100P100Q123943F

Package Exports

  • @wrud/cli

This package does not declare an exports field, so the exports above have been automatically detected and optimized by JSPM instead. If any package subpath is missing, it is recommended to post an issue to the original package (@wrud/cli) to support the "exports" field. If that is not possible, create a JSPM override to customize the exports field for this package.

Readme

@wrud/cli

What R U Doing — a local-first, open-source recorder for AI-agent sessions.

Your agent runs for an hour, changes dozens of files, spends real money, picks its own model — then the session scrolls off and is gone. wrud records every session (tools, models, tokens, cost, file edits, prompts), writes a plain-language recap, and turns recurring mistakes into lessons you feed back. Runs entirely on your machine — no cloud, no account, nothing leaves your box.

Works with Claude Code and Cursor via their own lifecycle hooks.

Quickstart

npx @wrud/cli

Starts the API + dashboard on one origin (default http://localhost:11190), seeds a local API key, opens your browser, and prints a token to paste on the Connect screen. On macOS it also puts a W in your menu bar (skip with --no-menubar).

Then wire your agent and verify capture end-to-end:

npx @wrud/cli install-hooks   # auto-detects & wires every agent you have (Claude Code, Cursor, ...)
npx @wrud/cli doctor

Restart your agent afterwards so it picks up the new hooks. (Target one agent with --agent claude-code / --agent cursor.)

npx @wrud/cli menubar

Installs a tiny native menu bar app (no Electron — the prebuilt universal .app ships inside this package) and launches it: a W icon that shows whether the server is running (dimmed = stopped), opens the dashboard, starts/stops the server, and shows today's sessions / tokens / estimated cost. Click Launch at Login in its menu to keep it across reboots.

Commands

Command What it does
wrud Start the server + dashboard (one origin). Attaches if already running.
wrud install-hooks [--agent claude-code|cursor] [--user|--project] Wire your hooks + mint a least-privilege ingest key + self-verify. No --agent auto-detects and wires every installed agent; --agent <id> targets one.
wrud doctor Prove the capture path works end-to-end (PASS/FAIL + HTTP status).
wrud menubar macOS: install + launch the native menu bar app (server start/stop, dashboard, today's usage).
wrud cleanup (alias uninstall) Remove everything wrud installed — ~/.wrud (db, tokens, log), temp session buffers, and wrud's hook entries in every agent's settings (user + project). Edits shared config surgically; --dry-run previews; confirms unless --yes.

Environment

Var Default Purpose
WRUD_PORT 11190 API/server port (dashboard is served same-origin)
WRUD_DB ~/.wrud/wrud.db SQLite database path
WRUD_BASE_URL http://localhost:11190 Base URL the hooks/CLI talk to
WRUD_API_KEY — Ingest token override (else read from ~/.wrud)
WRUD_NARRATOR_CMD claude CLI used for the plain-language recap (runs on your agent's own login; no API key)
WRUD_ANTHROPIC_KEY — Optional: use the Anthropic API for the recap instead of the local narrator

Privacy

Local-first by design: a Node server and one SQLite file on your disk. No account, no egress. The summary recap runs in the background on your agent's own login, so there's no extra API key and no extra bill.