How do I monitor Cursor agents?

Install ClawMetry (pip install clawmetry). The daemon auto-detects Cursor and renders every session, tool call, model, and token in one dashboard at http://localhost:8900. No IDE extension, no wrapper.

Which Cursor data does ClawMetry read?

Cursor's own local session history. ClawMetry reads it without an IDE extension and without a wrapper, and Cursor stays untouched.

Can I see when Cursor switches models mid-session?

Yes. Every turn is labelled with the model id. When a chat starts on Sonnet and switches to Opus for a hard turn, ClawMetry shows the switch and the cost delta. Roll-ups on the Cost tab show per-model spend per session.

Does ClawMetry work with Cursor's Composer, Chat, and Agent modes?

Yes. ClawMetry picks up all three identically. Composer runs (multi-file edits) get their tool-call sequence and per-turn tokens surfaced the same way as a normal chat.

How much does ClawMetry cost for Cursor?

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 per-run waste flags). Cursor is one of 18 paid runtimes on the Cloud tier.

Cursor × ClawMetry

Observability and cost tracking for Cursor agents

See every Cursor agent chat, tool call, and token as it happens. ClawMetry detects Cursor automatically and renders sessions, tokens, cost, and model routing in one dashboard. No IDE extension.

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

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

What ClawMetry sees on Cursor

Once ClawMetry is running, your Cursor 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: The token counts Cursor writes for each agent turn, normalised to input/output/cache-read/cache-write.
Cost per session, per model, per hour: cache-aware where the model supports caching. Compare Cursor spend against the other 19 runtimes ClawMetry observes.
Per-turn model routing (Cursor lets the user or the agent pick per turn between Claude Sonnet, GPT-4o, and others): the Cursor-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 Cursor 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 Cursor.

Start the free trial →

The one thing this catches that Cursor alone does not

Left alone, Cursor's failure mode is an agent chat that switches to Opus for one turn and stays there for the rest of the session. ClawMetry watches for it: cost per hour on the Cursor 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.