Keep an AI coding assistant consistent with your team's conventions across many separate chat sessions.
Store architectural decisions and past failed approaches so they are not repeated.
Generate context files like CLAUDE.md or .cursorrules automatically for tools that lack MCP support.
Enforce behavioral guardrails, like never touching certain files, before commits.
| anup-khanal/aidimag | ag-grid/ag-charts-server-side-example | caspermeijn/wallabag-test-server | |
|---|---|---|---|
| Stars | 0 | — | — |
| Language | Dockerfile | Dockerfile | Dockerfile |
| Last pushed | — | 2026-03-13 | 2024-12-24 |
| Maintenance | — | Maintained | Stale |
| Setup difficulty | easy | moderate | easy |
| Complexity | 2/5 | 2/5 | 2/5 |
| Audience | vibe coder | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires Node 18 or newer, optional semantic search needs an OpenAI key or local Ollama.
AI Dimag gives AI coding assistants like Claude, Cursor, Copilot, and Windsurf a persistent memory of your codebase that survives across separate chat sessions. Instead of re-explaining your project's decisions, conventions, and past mistakes every time you open a new session, this tool stores that knowledge as a set of claims saved in a folder called .aidimag right next to your code. What sets it apart is that each stored memory comes with evidence attached, such as a shell check, a linked commit, or a test result. A command called dim verify re-runs that evidence against your current repository, and if something that used to be true no longer holds, the memory is marked stale rather than quietly misleading the AI tool. Nothing gets added to memory automatically either: commits, pull requests, AI chat transcripts, and pasted documents get mined into proposals, but a human has to review and approve each one before it becomes an official memory. For tools that speak MCP, meaning Claude and Cursor, AI Dimag runs as a live MCP server offering search, proposal, and note-taking tools during a chat. For tools without MCP support, like Copilot and Windsurf, it can generate static context files such as .cursorrules, CLAUDE.md, and AGENTS.md that carry the same verified memory in a format those tools already read. Other features include guardrails, which are behavioral rules like never, ask first, or always that get enforced automatically, plus reusable step by step procedures called skills. There is also a team mode with a local server and sync command for sharing verified memory across a team, and a knowledgebase inbox that can summarize dropped in design docs or PDFs into memory entries. A web dashboard and editor extensions for VS Code and IntelliJ round out the tooling. Installation is a single npm command and requires Node 18 or newer. The project is free for teams of ten people or fewer under the Elastic License 2.0, with a commercial license required for larger or commercial use beyond that limit.
A memory system that lets AI coding tools remember your codebase's decisions and rules across sessions, with evidence that checks if those memories are still true.
Mainly Dockerfile. The stack also includes Node.js, npm, MCP.
Free to use for teams of ten or fewer people, larger teams or commercial use beyond that require a paid commercial license.
Setup difficulty is rated easy, with roughly 5min to a first successful run.
Mainly vibe coder.
This repo across BitVibe Labs
Don't trust strangers blindly. Verify against the repo.