Muse Code × ClawMetry
Muse Code usage monitor and cost tracking
See every Muse Code session, tool call and token as it happens. ClawMetry detects Meta's terminal coding agent automatically and renders sessions, transcripts, tokens and cost in one dashboard. No SDK, no config.
Muse Code, not the Muse app. Meta ships two products called Muse.
The consumer Muse personal agent runs entirely inside Meta's cloud on its
own secure VM: nothing of it exists on your machine, so no tool can observe it and
ClawMetry does not claim to. This page is about Muse Code, the
terminal coding agent, which runs locally.
All 31 runtimes ClawMetry observes
Muse Code is one of them; the other 30 are listed below. Free on every plan: OpenClaw, NVIDIA NemoClaw, Goose. Paid on Cloud Pro or Self-Hosted Pro: the rest.
Every runtime on one page, with what each one can and cannot show →
What ClawMetry sees on Muse Code
Once ClawMetry is running, your Muse Code activity shows up as:
- Sessions: started at, last activity, model, workspace, turn count, per-session token totals, and fork lineage when you branch a session.
- Full transcripts: user messages, the model's own reasoning, every tool call with its parsed arguments and output, your
! shell commands, sub-agents and compaction events.
- Tokens that are Muse's own numbers: ClawMetry reads the counted-once session totals Muse itself reports, not a reconstruction — input, output, cache reads and reasoning tokens, tagged with the model that produced them.
- Cache-aware cost: Muse Spark bills cached input at a fraction of the input rate, and coding sessions are cache-heavy. ClawMetry prices the cached slice at Meta's published cached rate and tells the two Muse Spark tiers apart, which differ more than tenfold.
- Cost next to the rest of your fleet: compare Muse Code spend against the other 30 runtimes ClawMetry observes, in one place.
The honest limits. ClawMetry reads Muse Code through Muse's own
session protocol rather than guessing at files, and it asks for nothing it does not
need: four read operations, and none of the capabilities that would let it run
commands in your workspace. Reading takes no lock, so your own Muse keeps running.
There is no pause, stop or kill for Muse Code sessions. Muse Code
runs a real process on your machine, so in principle it could be signalled —
but its protocol reports no process id, and ClawMetry will not ship a button whose
effect it cannot stand behind. The Guard tab says so plainly and offers the resume
command instead. That is a limit we expect to close, not a permanent one.
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
2. Run the daemon
clawmetry
The daemon finds Muse Code on its own and starts reading immediately. No env var, no config, no restart of Muse.
3. Open the dashboard
Open http://localhost:8900. Switch the runtime chip to Muse Code.
Start the free trial →
What this catches that Muse Code alone does not
A terminal agent with a million-token context is cheap per turn and expensive per
hour: the cost lives in long autonomous runs nobody is watching. ClawMetry gives you
cost per hour on the Muse Code runtime, repeated-tool and no-progress detection, and
per-session token growth — so a run that has stopped making progress is
something you find out about while it is still running.
Starred by the people building AI
Engineers and founders from OpenAI, Google, PostHog and more have starred ClawMetry on GitHub. One of them helped build it.
