A single Rust binary that gives AI agents durable cross-session memory as an MCP-native server — one binary, one SQLite file, no Docker, Postgres, or cloud. Exposes 40+ MCP tools spanning entity CRUD, hybrid search/RAG, an entity link graph, an immutable journal/audit trail, key-value state with TTL, and a memory lifecycle engine. Ships framework adapters for LangGraph, CrewAI, and AutoGen, plus a web dashboard.
- Storage
- Embedded SQLite with FTS5 for keyword search and stored dense vectors for semantic search; entities are idempotent by (category, key) with prior versions snapshotted into history for bi-temporal time-travel queries. Ebbinghaus decay scores age memories (buffer→working→core promotion, automatic archival, purge+VACUUM); optional transparent AES-256-GCM encryption of entity bodies (FTS index stays plaintext). Bundled quantized all-MiniLM-L6-v2 embeddings are compiled into the binary.
- Retrieval
- Hybrid recall by default: dense vector + FTS5 keyword fused via Reciprocal Rank Fusion, with Porter-stemming query expansion and graceful fallback to keyword search when no embeddings exist. Plus pure semantic search, bi-temporal `as_of` lookups, proactive just-in-time recall, entity-graph traversal, and `mimir_ask` RAG over stored memories via any Ollama/OpenAI-compatible LLM.
- Self-host
- Self-host: trivial
- License
- MIT
- Pricing
- Open-source MIT, free; one-line install of a single ~8MB binary with bundled embeddings (no network, no API key by default). No paid tier. · Free / OSS
- GitHub stars
- 9
- Last release
- 2026-06-28
- Last commit
- 2026-06-28
- First catalogued
- 2026-06-28
Strengths
- Zero-dependency local-first deployment: one Rust binary + one SQLite file, bundled in-process embeddings, fully offline with no API key
- Broad MCP surface (40+ tools): hybrid search/RAG, entity link graph, immutable journal/audit trail, KV state with TTL, and a decay-based memory lifecycle
- Permissive MIT license with security/compliance posture (AES-256-GCM at rest, air-gapped, published SBOM) and LangGraph/CrewAI/AutoGen adapters
Watch out
- Very new and low-adoption (~9 stars, created 2026-06); the comparison matrix and 100K-entity stress numbers in the README are self-reported and the project framing is marketing-forward
- Single-vendor (Perseus Computing LLC); long-term maintenance track record is not yet established
- Bundled embedder is a small all-MiniLM-L6-v2 model; higher-quality semantic recall needs an external Ollama/OpenAI embedding endpoint
Best for
- Local-first or air-gapped agents wanting a single-binary, MCP-native memory store with hybrid search, audit trail, and lifecycle decay
Benchmark results
No sourced results yet.
Sources
- Mimir README (vendor)
- GitHub API repo metadata (stars, MIT license, v2.7.0 release) (third-party)
Last verified 2026-06-28 · updated by discover-frameworks