Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
Version History
Critical fixes for memory indexing (amnesia), macOS binaries, root-start crashes, and massive indexing performance boost.
What This Skill Does
Muninn is a per-project memory layer for AI agents that uses the Context Exchange Protocol (CXP) to store and retrieve project context locally. All indexes, memories, and semantic search run on the host machine via a Rust-based engine. Memories are stored as plain Markdown files in `.muninn/` and can be inspected or edited directly.
Local-only processing means no code or context leaves the machine, unlike cloud-backed memory tools that send data to external APIs.
When to use it
- Persisting architectural decisions across agent sessions
- Searching a large codebase for existing patterns before writing new code
- Saving user preferences so agents don't repeat mistakes
- Orienting a new agent session to an unfamiliar project
- Rebuilding a stale project index after major refactors
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: 'Add a caching layer to this API project'
- 1Call brain_check to load the local project index and identify existing patterns
- 2Call search_context with 'caching' to find any prior caching decisions or utilities
- 3Review results to avoid duplicating existing work
- 4Implement the caching layer based on discovered patterns
- 5Call add_memory to record the architectural decision and cache strategy used
Caching layer added using existing project conventions, decision persisted locally for future sessions