Skip to content
StackUnderflow

StackUnderflow

Offline, local-first observability toolkit for AI coding agents. Ingests and indexes session logs from 17 coding agent providers to surface cost analytics, interactive session playback, and a searchable knowledge base.
Install View on GitHub

Quick start

Section titled “Quick start”
Terminal window
pip install stackunderflow
stackunderflow init

That opens the dashboard at http://127.0.0.1:8081.

What it does

Section titled “What it does”

StackUnderflow is an offline, local-first observability toolkit for AI coding agents. It ingests and indexes session logs from 17 coding agent providers to surface cost analytics, interactive session playback (with step-by-step filesystem reconstruction), and a searchable knowledge base that both developers and agents can query to learn from past decisions and failures. Everything runs locally with zero external dependencies or telemetry.

  • Dashboard — browse projects, sessions, token costs, and daily usage
  • Full-text search across every message you’ve sent or received
  • Q&A extraction — automatic question/answer pairs with code snippets
  • Bookmarks + auto-tags — save and categorise important sessions
  • CLI reports — stackunderflow today, month, optimize, export
  • SQLite-backed — incremental ingest, fast queries over hundreds of thousands of messages

The cost tab: spend by agent, cache ROI, and an error-cost breakdown by tool

Step-by-step playback with the reconstructed file tree at each moment

Ask StackUnderflow: a local model answering from your own session history

Where to next

Section titled “Where to next”