ContextSwitch is a kanban board designed for the way I actually work with AI coding agents. The core idea: drop a todo.db SQLite database in your repo and every agent (Claude Code, Cursor, Codex) can read and write to the same task list through MCP.
Features
- Shared Task Board: A single SQLite database that all your AI agents can access through MCP
- Visual Kanban: HTMX-powered web interface with drag-and-drop columns
- MCP Server: Exposes task management as MCP tools so agents can check status, claim tasks, and mark things done
- Project Diaries: Automated development journals that track decisions, progress, and blockers using Claude Code hooks
- Lexicon Integration: Personal prompt library that turns your templates into MCP slash commands
How It Works
When you’re juggling multiple AI agents on a project, the hard part is coordination. ContextSwitch gives them a shared state by putting the task database right in the repo. Start the MCP server, and agents can see what needs doing, what’s in progress, and what’s blocked.
The diary feature hooks into Claude Code so every coding session automatically logs what happened: what changed, what decisions were made, what’s still open. It’s like having meeting notes for your AI pair programming sessions.