Install & run — one command

npx agentgraphed

That's the whole install. No clone, no signup, no config file. The dashboard opens at http://localhost:3737 the moment it's ready. Re-run any time to scan for new sessions.

Prefer a global install?

npm install -g agentgraphed
agentgraphed

Don't have npx / Node? Install Node 20+ from https://nodejs.org (the LTS version is fine). Check yours with node --version.

What to expect on first run

1. npx agentgraphed downloads the package once (~15 MB, takes 30-60s on a typical connection). Subsequent runs start in seconds — npm caches the package.
2. You'll see one npm warn exec line and a couple of npm warn deprecated warnings from transitive dependencies. These are harmless; they'll go away in a future release.
3. The dashboard server boots. You'll see:

   › Starting AgentGraphed on http://localhost:3737
   › Scanning local AI coding sessions…
     Found 142 Claude + 8 Codex sessions (14821 messages indexed in 1943ms)
   › Ready. Press Ctrl+C to stop.
   

4. Your browser opens to the dashboard automatically.
5. To stop the server, hit Ctrl+C in that terminal. Your indexed data stays at ~/.agentgraphed/agentgraphed.sqlite for next time.

What it does

AgentGraphed reads your local Claude Code (~/.claude/projects/) and Codex CLI (~/.codex/sessions/) JSONL logs and turns them into a real dashboard. There's no agent to install on each session, no API key required to get started — it just reads files your CLI tools were already writing.

After npx agentgraphed, here's how to use it:

• Browse your work — Timeline groups every session by day. Multi-day sessions show STARTED · SPANS Nd / CONTINUED / CLOSED badges so nothing hides on a single bucket. Click any session to read the full conversation in a chat-bubble view.
• Resume a session — On any session page, click Resume session to copy cd <cwd> && claude --resume <id> to your clipboard. Paste it in your terminal to pick up where you left off.
• Find a specific project — Projects ranks every git repo you've worked in by activity, tokens, and cost. Click one to see only that repo's sessions.
• Refresh after more coding — Re-run npx agentgraphed. It re-scans your CLI logs incrementally; already-ingested sessions are skipped.
• Get clean titles and categories (optional, BYO key) — Open Settings → LLM provider, paste an Anthropic or OpenAI key, then click Classify uncategorized. Past-tense titles like "Fixed Stripe checkout bug" replace the raw first prompt. Cost is typically $0.01-0.03 for a few hundred sessions.
• Live quota probe (optional, opt-in) — On the dashboard, click Probe now (or toggle Poll every 60s) for live 5h & 7d rate-limit utilization read straight from Anthropic. Each probe costs a single token (~$0.00006).

Features

• Dashboard — 30-day usage chart, KPIs, top projects, work categories at a glance
• Timeline — every session grouped by day, searchable, filterable by project or provider
• Projects — auto-detected from git repo roots; see which projects pull the most AI time
• Sessions — read past conversations in a clean chat-bubble view
• Resume — one click copies cd <cwd> && claude --resume <id> to your clipboard
• Copy context (optional, BYO LLM key) — generates a primer for a fresh chat
• Multi-label classification (optional, BYO LLM key) — auto-categorizes sessions as Feature / Debugging / Planning / Refactor / Styling / DevOps / Data / Payments / Docs / Content
• Cost estimates — uses LiteLLM's auto-updating pricing data (2700+ models covered)
• Range picker — 7d / 30d / 90d / all-time on every chart

Screenshots

Timeline — every session, grouped by day. Multi-day sessions get STARTED · SPANS Nd / CONTINUED / CLOSED badges so nothing hides on one bucket.	Session detail — read past conversations in a clean chat-bubble view. One click to resume in Claude Code or copy a primer for a fresh chat.
Projects — every git repo you've worked in, ranked by activity. See which projects are eating your week.	Analytics — sessions per day, provider split, model breakdown with cost per model.

Privacy

• Everything lives in ~/.agentgraphed/agentgraphed.sqlite on your machine.
• Nothing is uploaded unless you click "Copy context" or "Classify sessions," and even then only to your LLM provider with your API key.
• API keys are stored in plaintext in the SQLite file — same threat model as ~/.aws/credentials or a .env. Don't commit your home folder to git.

Supported sources

• Claude Code — reads ~/.claude/projects/*/*.jsonl
• Codex CLI — reads ~/.codex/sessions/YYYY/MM/DD/*.jsonl

More adapters coming. If you want a specific one, open an issue.

Optional: LLM session titles & categories

Without an API key, sessions show the first line of the first prompt as the title. With a key:

• Anthropic — Haiku 4.5 (default) gives clean past-tense titles like "Fixed Stripe checkout double-decimal bug"
• OpenAI — GPT-5 mini works equally well, often cheaper

Click Settings → LLM provider, paste your key, then Classify uncategorized. Cost is fractions of a cent per session — typically $0.01–0.03 for a few hundred sessions.

Configuration

Environment variables:

Log directories can also be edited at Settings → Data sources without restarting.

Development

git clone https://github.com/sudomichael/agentgraphed.git
cd agentgraphed
npm install
npm run dev

Open http://localhost:3737. Hot reload, source under src/.

To produce a publishable build:

npm run build         # builds Next.js standalone bundle
npm pack              # creates agentgraphed-X.Y.Z.tgz

License

MIT © Michael Patrick