muninn

Verified·Scanned 2/17/2026

Universal Context Protocol (CXP) for AI agents. Provides per-project long-term memory, semantic search, and surgical context injection. Features a mandatory 'Brain-First' Gate Keeper (v2.1.4) that prevents hallucinations. Created by @Blackknight1dev.

from clawhub.ai·v1.6.0·2.4 KB·0 installs
Scanned from 1.6.0 at baadc85 · Transparency log ↗
$ vett add clawhub.ai/endgegnerbert-tech/muninn

Muninn - The Universal Memory Layer

Muninn is an advanced per-project memory system designed for AI agents. It uses the Context Exchange Protocol (CXP) to provide high-performance, token-efficient context management.

🛡️ Integration

Muninn operates as an MCP (Model Context Protocol) server. To use this skill, ensure that the muninn-core package is installed in your environment.

🚀 The Muninn Protocol (v2.1.4 - ENFORCED)

As an agent, you SHOULD follow these three pillars of the Muninn workflow to remain efficient and accurate. Note that the MCP server blocks non-exempt tools if brain_check hasn't been called first in the current session.

  1. Orient Yourself (brain_check): Call this at the start of every session or when switching projects. It loads the project index and identifies key files and patterns.
  2. Search Before You Code (search_context): Never assume you know the codebase. Search for existing patterns or prior decisions to avoid duplication and bugs.
  3. Learn as You Go (add_memory): Every significant fix, architectural decision, or user preference should be saved as a memory. This ensures persistence across future sessions.

🛠 Tool Reference (MCP)

  • init_project: Initialize Muninn in a new project directory.
  • brain_check: MANDATORY first step. Loads context and patterns.
  • add_memory: Persist facts, decisions, and lessons learned.
  • search_context: Semantically search project code and memories.
  • reindex_context: Rebuild the semantic index to reflect file changes.
  • health_check: Verify the status of the Muninn system.

🦞 Pro-Tip for Agents

Muninn surgically retrieves only what's needed. Instead of reading large files, use search_context to find the exact lines that matter. This significantly reduces context window overhead and token costs.