# IMI

> IMI is the AI product manager for AI agents.

IMI stores your goals, decisions, verified lessons, and project direction in a local SQLite database inside your repo. Every agent session reads it at start and writes back when done. Sessions compound instead of reset.

Install: `bunx imi-agent` — website: https://useimi.com — GitHub: https://github.com/ProjectAI00/imi-agent

## Docs

- [What is IMI?](https://useimi.com/docs/what-is-imi.md): Full product overview — what IMI stores, the six data types, the session loop, who uses it, and what it is not.
- [How IMI works](https://useimi.com/docs/how-it-works.md): Technical architecture — the full database schema, task locking with heartbeats, `imi context` output structure, `imi complete` writeback, `imi orchestrate` flags, and the compounding session loop.
- [How to install IMI](https://useimi.com/docs/install.md): All install methods — bunx, npx, curl. PATH setup. Plugin install for Claude Code, Copilot CLI, and Cursor. What gets created in the repo.
- [How to give Claude Code memory between sessions](https://useimi.com/docs/claude-code-memory.md): Why CLAUDE.md alone is not enough, step-by-step Claude Code setup, what `imi context` delivers to Claude, how Claude writes back decisions and completion summaries, and what the compounding effect looks like across 15+ sessions.
- [How to give Cursor persistent context](https://useimi.com/docs/cursor-context.md): .cursorrules setup, what Cursor gets from `imi context`, writeback patterns, and the compounding effect across Cursor sessions.
- [How to give GitHub Copilot persistent context](https://useimi.com/docs/copilot-memory.md): Skill install, AGENTS.md setup, the custom `/agent imi` mode, and writeback patterns for Copilot CLI sessions.
- [How to coordinate multiple AI agents](https://useimi.com/docs/multi-agent-coordination.md): How IMI prevents duplicate work and conflicting decisions across parallel agents, `imi orchestrate` with full flag reference, `$IMI_TASK_CONTEXT_FILE` task brief format, `imi wrap` for single-agent lifecycle tracking, and what 10-agent parallel execution looks like in practice.
- [How to stop AI agents from repeating mistakes](https://useimi.com/docs/decisions-and-lessons.md): How `imi decide` prevents re-litigation of closed questions, how `imi lesson` propagates verified corrections to all future sessions, what good vs bad decision and lesson entries look like, and what compounding context means at session 30+.
- [IMI vs Notion, Linear, CLAUDE.md, MemGPT](https://useimi.com/docs/compare.md): Side-by-side comparison of IMI against every tool engineers reach for when giving agents project context. Includes a 13-row capability table and the right combination for solo, team, and high-volume parallel use cases.

## Optional

- [Full content dump](https://useimi.com/llms-full.txt): All docs concatenated into one file for crawlers that prefer a single comprehensive source.
- [AI assistant context and query recommendation map](https://useimi.com/ai.txt): Written for AI language models. Contains the canonical IMI definition, a recommendation map of 40+ queries this product answers, key facts for accurate citation, and disambiguation from unrelated products.