How do I monitor Pi agents?

Install ClawMetry (pip install clawmetry). The daemon auto-detects Pi and renders every session, tool call, model, and token in one dashboard at http://localhost:8900. No SDK, no code change to Pi.

Does ClawMetry work when Pi routes across Anthropic, OpenAI, and local models?

Yes. Pi's per-turn model routing writes the model id on every turn; ClawMetry reads it and shows the routing sequence per session, plus per-model spend rolled up on the Cost tab. You can see when Pi flipped from a local model to a paid API and what that cost you.

Can I see Pi cost per hour and per session?

Yes. The Cost tab computes per-session, per-model, per-hour spend from Pi's turn-level token counts. Set a per-hour budget alert on the Pi runtime and the approval queue pauses the session before the next call if it crosses the threshold.

Does ClawMetry track Pi's sub-agents?

Yes. When a Pi session spawns a child, ClawMetry shows the tree with per-child token totals and tool calls rolled up under the parent turn. Same shape as Claude Code's Task tool trees.

How much does ClawMetry cost for Pi?

Free forever on your own machine. ClawMetry Cloud starts at $9 per node per month (Starter) or $19 per node per month (Pro, which adds policy gates and evals). Pi is one of 18 paid runtimes on the Cloud tier.

Pi × ClawMetry

Observability and cost tracking for Pi agents

See every Pi agent session, tool call, and token as it happens. ClawMetry detects Pi automatically and renders sessions, tokens, cost, and model routing in one dashboard. No SDK.

Supported runtime Zero config, auto-detected Read-only Free trial Official Pi site ↗

📦 631k+ installs 🌍 128+ countries ⭐ 392 on GitHub 🏆 #5 Product Hunt

What ClawMetry sees on Pi

Once ClawMetry is running, your Pi activity shows up as:

Sessions: started at, last activity, model, message count, per-session token totals.
Live activity: every tool call as it lands, with args and result. Turn-level tokens split.
Per-turn tokens: Per-turn token counts written by Pi's runtime (input, output, and cache-read tokens when applicable).
Cost per session, per model, per hour: cache-aware where the model supports caching. Compare Pi spend against the other 19 runtimes ClawMetry observes.
Model-agnostic routing (Pi swaps between Anthropic, OpenAI, and local models turn-by-turn): the Pi-specific signal ClawMetry surfaces that generic dashboards do not.
Approval + kill switch: watch for token-maxing signals (message repeat, no-progress heuristic, cost-per-hour threshold), pause or kill the session before the next call.

Enable it in about two minutes

1. Install ClawMetry

Pick the line that matches your OS. Windows users: paste the install.cmd line, not the shell one.

# macOS or Linux
curl -fsSL https://clawmetry.com/install.sh | bash

# Windows (cmd.exe or PowerShell)
curl -fsSL https://clawmetry.com/install.cmd -o install.cmd && install.cmd

Already have Python? pip works too; onboard is the interactive first-run wizard.

# Any OS with Python 3.9+
pip install clawmetry && clawmetry onboard

Or skip the terminal entirely and grab the desktop app: macOS (.dmg) · Windows (.exe) · Linux (.AppImage). It installs and updates ClawMetry for you.

2. Run the daemon

clawmetry

The daemon auto-detects Pi and starts reading immediately. No env var, no config, no runtime restart.

3. Open the dashboard

Open http://localhost:8900. Switch the runtime chip to Pi.

Start the free trial →

The one thing this catches that Pi alone does not

Left alone, Pi's failure mode is a session that starts on a local model and silently falls back to a paid API on every hard turn. ClawMetry watches for it: cost per hour on the Pi runtime, message-repeat detection, and per-session token-growth curves. Set a threshold, and the approval queue pauses before the next call. Kill switch (Pro) stops it.