npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2026 – Pkg Stats / Ryan Hefner

clawdeck-panel

v0.3.0

Published

Unofficial local dashboard for Claude Code: sessions, events, cost, worktrees, reviews - zero dependencies, loopback-only.

Readme

Clawdeck

CI License: MIT Node Dependencies GitHub stars

An unofficial local dashboard for Claude Code. Point it at any project you work on with Claude Code and it shows what is actually happening: live sessions, an event timeline, cost and context telemetry, git worktrees, reviews, and delivery state - in one local web UI.

The Clawdeck overview dashboard, dark theme

Feature tour

Trace waterfall - every turn of a session broken into tool-call spans with real durations: subagent tasks, failing commands, and human-wait spans (dashed, width-capped) at a glance.

Trace waterfall: turns with tool-call spans on a timeline

Burn rate & forecast - $/hour from statusline cost deltas, 5h/7d depletion slopes with ETA, per-model token history over 7d/30d/all-time windows. Estimates are labelled as estimates; unknowns stay unknown.

Cost hub: burn KPIs, forecast card, per-model history

Ask Clawdeck (Prompt hub) - ask questions about panel state, answered by a local claude -p child running tool-less in a sterile temp dir; the only context sent is a compact, secret-scanned snapshot summary.

Also in the box:

  • Config map (Configuration) - every rule, slash command, skill, agent, MCP server, and hook the checkout declares to Claude Code, overlaid with what recent sessions actually invoked. Dead config shows up dim.
  • MCP & skills analytics (Cost) - per-server call counts, error rates, and durations from recent transcripts: evidence for whether a server earns its context cost.
  • Host vitals (Health) - CPU, memory, and checkout-volume disk next to the panel's own self-performance numbers.
  • Cheap refreshes - snapshots carry per-section content hashes, so unchanged views skip re-rendering, /api/snapshot answers 304, and the page revalidates when you come back to the tab.

Principles:

  • Zero dependencies. Pure Node stdlib on the server, browser-native ES modules in the UI. No build step, no node_modules.
  • Loopback-only. Binds 127.0.0.1, refuses foreign Host headers, gates privileged routes behind a per-launch bearer token.
  • Degrades gracefully. Everything works read-only on a bare git repo; more signal appears as you opt in to the hooks, statusline bridge, and OTEL.

Clawdeck is a community project. It is not affiliated with or endorsed by Anthropic.

Quickstart

npx clawdeck-panel run --checkout /path/to/your/project

Or from a clone:

git clone https://github.com/m-sanchez/clawdeck.git
cd clawdeck
node scripts/panel-run.mjs --checkout /path/to/your/project

That alone gives you the git-level views (worktrees, diff, commits, MR draft) and session liveness from Claude Code's own transcript files — zero setup, nothing written to your project, one loopback server that stops when you close it.

npx vs a clone for init. The one-off run above is fine over npx. But init (below) writes generated /panel slash commands that reference the panel's install path — under npx that is the npm cache directory, which npm may garbage-collect. If you plan to keep the integration installed, run init from a clone (or a global install) so the referenced path is stable.

Install the integration (optional, recommended)

The event timeline, activity feed, and cost views are fed by a tiny hook + statusline bridge you install into the observed project:

npx clawdeck-panel init --target /path/to/your/project --statusline

init copies the emitter hook (and its lib) into the project's .claude/hooks/, installs /panel slash commands, and prints the hook registrations to paste into .claude/settings.json - or merges them for you with --write-settings (only ever appends its own entries; backs up first). Restart your Claude Code session afterwards.

What you get at each level

| Setup | What lights up | | ----------------------- | ----------------------------------------------------------------- | | bare git repo | Overview, worktrees, diff/review views, MR draft, session pulse | | + event hooks | live event timeline, per-session activity, delivery lifecycle | | + statusline bridge | live cost, context-window, and model telemetry per session | | + OTEL exporter pointed at the panel | token/cost metrics via OTLP-JSON | | + forge token | MR/PR + pipeline status, merge tracking, notifications |

To point Claude Code's OTEL exporter at the panel, set these before launching Claude (the panel prints its port and token on start; the /v1/metrics endpoint requires the panel token):

export CLAUDE_CODE_ENABLE_TELEMETRY=1
export OTEL_METRICS_EXPORTER=otlp
export OTEL_EXPORTER_OTLP_PROTOCOL=http/json
export OTEL_EXPORTER_OTLP_ENDPOINT=http://127.0.0.1:<panel-port>
export OTEL_EXPORTER_OTLP_HEADERS=x-panel-token=<panel-token>

Forge connectors

Clawdeck auto-detects the project's git host from origin and speaks to it read-only:

  • GitHub (github.com + GHES) - PRs and Actions runs. GITHUB_TOKEN optional for public repos.
  • GitLab (gitlab.com + self-hosted) - MRs and pipelines. Needs GITLAB_TOKEN.
  • Bitbucket Cloud - PRs and Pipelines. Needs BITBUCKET_TOKEN (a repository/workspace access token).
  • Azure DevOps - PRs and builds. Needs AZURE_DEVOPS_TOKEN (a PAT).
  • Gitea / Forgejo - PRs and commit status. Self-hosted hosts are anonymous, so opt in with GITEA_URL (+ GITEA_TOKEN for private repos).

Tokens live in the observed project's .claude/settings.local.json (or env) and never reach the browser.

Roadmap

  • Session → subagent hierarchy tree view.

Architecture

Claude Code hooks ──► durable spool (at-least-once) ──► single-writer store
                                                            │
statusline bridge ──► per-session telemetry records ────────┤
OTLP-JSON exporter ─► OTEL receiver ────────────────────────┤
                                                            ▼
                                             HTTP + SSE server (loopback)
                                                            ▼
                                             browser SPA (no build step)

The server never watches the filesystem; it polls adapters per request and on a bounded SSE interval. See ARCHITECTURE.md and docs/DECISIONS.md.

Security model

  • Loopback bind + strict Host allowlist (anti-DNS-rebinding).
  • Per-launch bearer token, delivered to the browser only in the URL fragment (never in served HTML, API bodies, logs, or Referers).
  • Separate per-launch ingest token for the event POST route.
  • PID+nonce ownership checks so lifecycle scripts can never kill a reused PID.
  • Single-writer lock on the canonical event store; a second panel degrades to its own local store instead of corrupting the shared one.
  • Deep links to Claude are fail-closed secret-scanned: a prompt containing suspected secret material refuses to become a URL.
  • No shell endpoint. Commands are a fixed allowlist with server-built argv.

Details: docs/SECURITY.md.

Development

npm test              # node --test
npm run self-test     # boots the server against this repo and checks /health

License

MIT