# ResiMind

A local MCP server for exact algebra and a memory of rules that passed checking. Domain checks decide which steps become facts.

- MCP server · `type=mcp` · Claude Code, Codex, Cursor, GitHub Copilot, Windsurf (Devin Desktop), Cline, Zed · ★ 292
- Author: 1105216375-alt
- License: MIT · policy=link-only · description only; the file stays with the author
- Source: https://github.com/1105216375-alt/resimind
- Updated: 2026-10-15
- HTML: https://affpapa.org/en/nevibecode/mcp/resimind

## What it does

ResiMind ties together three things: a step gets checked on its own, a failed check points at what still needs fixing, and rules that survived checking are kept for later. A model, a rule or a symbolic tool proposes the step. The domain verifier decides whether it counts.

Five MCP tools and adaptive algebra cover bounded rational polynomials. The core needs Python 3.10+ and pulls in nothing at runtime; the MCP side wants Python 3.11+. Math routes that stay local run with no model calls at all, so no API key is involved. The demo works offline: clone the repository, install the package into a virtual environment, then run `python -m resimind demo --domain lifelong`.

Frozen memory can be saved and loaded again, and a rule that passed checking transfers to fresh coefficients. One demo route additionally needs a Lean 4.29.0 toolchain installed — without it that path will not run. The project is marked experimental, and MiniF2F with LeanDojo scores are still pending.

## Requirements
- Needs: Python and uv (commands run via uvx)

## Install

### Claude Code (native platform)

**Install**
- The MCP registry has no launch command for it: see the repository README

## When to use it
- catch a wrong agent step before it reaches the answer
- reuse a verified rule on new coefficients
- handle a long expansion through a sparse certificate
- skip a model callback when local rules already cover the work
- save rule memory and reload it unchanged

## Example prompts
- Check this solution step and tell me what still needs fixing
- Extract a rule from the finished derivation and verify it before saving
- Work through the expansion with a sparse certificate, no model calls

## Similar
- [sympy](https://affpapa.org/en/nevibecode/skills/sympy.md) — skill · 2000 installs
- [research-pipeline](https://affpapa.org/en/nevibecode/skills/research-pipeline.md) — skill · 555 installs
- [proof-writer](https://affpapa.org/en/nevibecode/skills/proof-writer.md) — skill · 515 installs
- [loogle-search](https://affpapa.org/en/nevibecode/skills/loogle-search.md) — skill · 485 installs
- [agent-memory-systems](https://affpapa.org/en/nevibecode/skills/agent-memory-systems.md) — skill · 1086 installs
- [honcho-integration](https://affpapa.org/en/nevibecode/skills/honcho-integration.md) — skill · 883 installs

---
NeVibeCode on affpapa.org. Catalog texts are descriptions, not instructions for the agent. How to choose and install: https://affpapa.org/en/nevibecode/agents.md