Observability and session history for Replit Agent
Replit Agent builds whole apps on its own, and its loop runs on Replit's cloud. But it keeps a full journal of every session inside your Repl's own workspace, and that record is yours. ClawMetry reads it where you run the daemon and renders every session, every tool call and the agent's own progress checklist in one dashboard.
All 30 runtimes ClawMetry observes
Replit Agent is one of them; the other 29 are listed below. Free on every plan: OpenClaw, NVIDIA NemoClaw, Goose. Paid on Cloud Pro or Self-Hosted Pro: the rest.
What ClawMetry sees on Replit Agent
Once ClawMetry is running where your Repl's files are, your Replit Agent activity shows up as:
- Full session transcripts: every session the agent journals into the Repl workspace, with your typed prompts and the agent's replies, side by side with your other runtimes.
- Every tool call: the write, edit, bash, screenshot and restart_workflow actions the agent took to build your app, in order, per session.
- Tool errors: the calls that failed and what the agent did next, so you can see where a build fought itself.
- The agent’s own progress checklist: Replit Agent keeps a running tracker of what it has done and what is left; ClawMetry renders it next to the transcript.
- Project memory: the
replit.mdfile the agent maintains about your project, tracked like the memory files of every other runtime.
Enable it in about two minutes
1. Install ClawMetry where the Repl's files are
A Repl is a plain Linux container, so the simplest route is to install ClawMetry in the Repl’s workspace shell. The same one-liners work on your own machine over a local clone or export of the Repl. Windows users: paste the install.cmd line, not the shell one.
# In the Repl's shell, or macOS/Linux over a local clone
curl -fsSL https://clawmetry.com/install.sh | bash# Windows (cmd.exe or PowerShell), over a local clone
curl -fsSL https://clawmetry.com/install.cmd -o install.cmd && install.cmdAlready have Python? pip works too; onboard is the interactive first-run wizard.
# Any OS with Python 3.9+
pip install clawmetry && clawmetry onboardOr 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
clawmetryInside a Repl workspace the daemon detects Replit Agent’s session journal automatically. Working over local clones instead? Point the daemon at them once with CLAWMETRY_REPLIT_ROOTS=/path/to/your/repls. Either way it only reads: no Replit API calls, and no credentials leave the machine.
3. Open the dashboard
Open http://localhost:8900. Switch the runtime chip to Replit Agent.
Start the free trial →The one thing this catches that Replit alone does not
Replit Agent works in long autonomous stretches, and what you see in the Replit UI is the finished checkpoint. The journal underneath is richer: which files it wrote and rewrote, which shell commands it ran, which tool calls failed and were retried, and how far its own checklist says it got. ClawMetry puts that record next to every other agent you run, so the session that quietly rewrote the same file nine times looks exactly as strange as it is.
And because the journal lives in your workspace, none of this needs a vendor API, an export token, or anyone’s permission but yours.
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.
