wolfiesch

Pi Trajectory — Development skill for Claude Code

Development community

Investigate and replay Pi-compatible coding-agent sessions across agents, messages, tools, files, tokens, costs, waits, and failures.

How to install Pi Trajectory

This entry records only its repository, not the path inside it, so there is no exact command to give. Open wolfiesch/pi-trajectory and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Pi Trajectory does

Investigate and replay Pi-compatible coding-agent sessions across agents, messages, tools, files, tokens, costs, waits, and failures.

Alternatives in Development

Code Review Graph — Local knowledge graph for Claude Code 3.7k ★
Claude Supermemory — Persistent memory across sessions and projects using Supermemory 2.3k ★
Pi Skills — Skills for pi coding agent (compatible with Claude Code and Codex CLI) 1.3k ★

README

Pi Trajectory

[![CI](https://github.com/wolfiesch/pi-trajectory/actions/workflows/ci.yml/badge.svg)](https://github.com/wolfiesch/pi-trajectory/actions/workflows/ci.yml) [![MIT License](https://img.shields.io/badge/license-MIT-blue.svg)](LICENSE)

Local observability for Pi-compatible coding agents: what every agent in a session did, said, spent, and touched, as a scrubbable, replayable dashboard. The IRC direct messages between subagents are first-class signals, not a footnote.

It reads the local transcripts OMP and upstream Pi already maintain, so there is nothing to instrument and no agent code to change. A finished session can be scrubbed and replayed; a running session streams.

![Pi Trajectory overview showing investigation findings, run filters, session details, and playback controls](assets/pi-trajectory-overview.png)

Run

For macOS or Linux with [Bun](https://bun.sh/) installed:

git clone https://github.com/wolfiesch/pi-trajectory.git
cd pi-trajectory
bun install
bun run dev          # API on :8787, UI on http://localhost:5273
# or serve the bundled synthetic fixture instead of local sessions:
bun run bin/pi-trajectory.ts --demo

Separately, if you prefer:

bun run api          # transcript parser + REST + SSE
bun run web          # Vite dev server, proxies /api

Freeze one session for offline work, or export a privacy-safe replay bundle:

bun run fixture                       # newest session that dispatched subagents
bun run fixture my-project            # newest match on project, title, or agent
bun run fixture ''       # exact key from /api/sessions
bun run bundle ''        # redacted replay graph in bundles/
bun run bundle '' --out /tmp/session.json