aidebrief 0.2.1

Passive AI conversation capture, full-text search, and decision extraction for your local dev environment

aidebrief

Passive AI conversation capture + full-text search + decision extraction for your local dev environment.

Every AI coding conversation — from Copilot, Claude Code, opencode — vanishes the moment you close the session. aidebrief captures them automatically, extracts engineering decisions, and makes everything searchable in a local SQLite database. No cloud, no telemetry, no infrastructure.

Quick start

pip install aidebrief
aidebrief init
aidebrief search "why did we use sqlite instead of postgres"

How it works

Hook files from Copilot and Claude Code are written to disk when a conversation ends. On your next search, context, or supervise call, SyncManager processes them lazily — ingesting the transcript, indexing it with FTS5, and extracting structured engineering decisions. No background daemon needed.

An MCP server exposes the database via tools that AI agents can call directly. Opencode sessions can be imported manually via aidebrief import-opencode --days 30.

Features

• Capture — Hooks into Copilot, Claude Code, opencode, and git
• Search — Full-text search across all conversations
• Context — Given your current question, returns the most relevant past sessions
• Decisions — Automatically extracts structured engineering decisions
• Contradictions — Checks if a new proposal conflicts with past decisions
• Supervise — One combined call returning context + decisions + contradictions — designed for AI subagents

Requirements

Python 3.11+ and mcp>=1.0.0. No other dependencies.