Skip to main content

​
Overview

The Pi extension traces every local Pi coding agent session: each prompt becomes a full trace with its LLM calls and tool executions, sent to Laminar. Traces from the same Pi session are grouped into one Laminar session. The extension subscribes to Pi’s own lifecycle events, so spans open and close in process. Traces appear in Laminar while the run is still going.

​
Install

1

Run the installer

The CLI logs you in (browser device flow), lets you pick the Laminar project that should receive your Pi traces, mints a project API key named after the extension and your machine, writes it to ~/.config/lmnr/pi-extension.json, and installs the extension with pi install npm:@lmnr-ai/pi-extension.
Pick a dedicated project for coding-agent traces so they don’t mix with your application’s traces. The setup is global and directory-independent: it never touches .lmnr/project.json or your .env.
2

Restart Pi

Pi loads extensions at startup, so restart it to activate the extension.
3

Use Pi as usual

Every prompt becomes a trace in the project you picked. Open the Laminar dashboard to watch them land in realtime.
Self-hosting Laminar? The same command works against your instance: pass your frontend URL (the auth issuer) and your API URL, including the API port:

​
Manual install

If you’d rather not use lmnr-cli, install the package with Pi’s own command and set the key in your environment:
Get a key from the Laminar dashboard under Settings → Project API Keys (create a dedicated key so it’s clear it belongs to the extension). pi install adds the package to your global settings (~/.pi/agent/settings.json) and installs it under ~/.pi/agent/npm/. Add -l to scope it to the current project (.pi/settings.json) instead. You can also add the package to settings.json by hand:
~/.pi/agent/settings.json

​
What you get

One trace covers one prompt. Pi can run its agent loop more than once for a single prompt: an auto-retry, an auto-compaction retry, and a queued follow-up each start another pass. Every pass lands in the same trace. Inside it you get one LLM span per model call and one tool span per tool call. Spans open when the work starts, not when it finishes. An LLM span opens as Pi hands the request to the model, so its duration is the real call latency.

​
Configuration

The extension reads its config from ~/.config/lmnr/pi-extension.json, written by lmnr-cli plugin add pi:
projectApiKey is required. baseUrl is optional and defaults to https://api.lmnr.ai; self-hosters set it to their instance’s API URL (for example http://localhost:8000). These environment variables override the file when set, which is handy for CI or a shell you already have configured. The extension is fail-open: with no key from either source it disables itself and Pi runs untouched. Traces are attributed to the identity from lmnr-cli login when present; set LMNR_USER_ID to override the user id attached to traces.

​
Debugger sessions

Set LMNR_DEBUG=true and each run also joins a Laminar debugger session, so you can inspect and replay it. The SDK resolves the session id in this order: LMNR_DEBUG_SESSION_ID, then the nearest .lmnr/debug-session.json written by lmnr-cli debug session new, then a fresh id. LMNR_DEBUG=true also turns on file logging to ~/.pi/agent/lmnr-pi-extension.log. Pi is a terminal UI, so the extension never writes diagnostics to stdout. Read that file when traces do not arrive.