Show HN: Mimir – local-first encrypted memory for AI agents (single Rust binary)

Perseus Computing LLC released Mimir, a local-first, encrypted memory engine for AI agents as a single Rust binary with zero dependencies. The open-source tool provides persistent memory across sessions via the Model Context Protocol, supporting 46 tools and AES-256-GCM encryption, and works with clients like Claude Desktop and Cursor. Mimir aims to differentiate from competitors like Mem0 and Letta by being fully offline, MCP-native, and dependency-free.

Persistent Memory for AI Agents — MCP-Native. Local-First. Zero Dependencies. Mimir is a single Rust binary that gives AI agents durable memory across sessions. One binary. One file. No Docker. No Postgres. No cloud. Just persistent memory that works with any MCP host. curl -sSf https://raw.githubusercontent.com/Perseus-Computing-LLC/mimir/main/scripts/install.sh | sh That's it. Mimir is installed to ~/.local/bin/mimir . Start it: mimir serve --db ~/.mimir/data/mimir.db macOS note.On Apple Silicon, an unsigned binary is killed on launch Killed: 9 , no output by the OS binary policy — even with no quarantine attribute. The installer ad-hoc code-signs Mimir for you. If you build or copy the binary yourself cargo build --release && cp target/release/mimir ~/.cargo/bin/ , sign it once after each rebuild: codesign --sign - "$ command -v mimir " Connect any MCP host Claude Desktop, Cursor, Hermes Agent, Perseus, etc. : { "mcpServers": { "mimir": { "command": "mimir", "args": "serve", "--db", "~/.mimir/data/mimir.db" } } } Start Mimir mimir serve --db memory.db & sleep 1 Remember a fact via MCP JSON-RPC on stdio echo '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"mimir remember","arguments":{"category":"demo","key":"hello","body json":"{\"text\":\"Hello from Mimir \"}"}}}' | mimir serve --db memory.db Search for it echo '{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"mimir recall","arguments":{"query":"Hello"}}}' | mimir serve --db memory.db Mimir is a standard MCP stdio server — the same mimir serve command works everywhere. Run mimir doctor to validate your install and print this matrix locally. | Client | Status | Config | |---|---|---| | Claude Desktop | ✅ | claude desktop config.json | | Claude Code / Hermes | ✅ | .mcp.json / config.yaml | | Cursor | ✅ | .cursor/mcp.json | | Windsurf | ✅ | mcp config.json | | VS Code + Continue.dev | ✅ | config.json | | Zed | ✅ | settings.json | | Codex CLI | ✅ | ~/.codex/config.toml | Copy-paste config snippets for each: docs/clients/ . Mimir is the only memory engine that is simultaneously MCP-native, local-first, zero-dependency, AND agent-first. | Mimir | Mem0 | Letta | Zep | | |---|---|---|---|---| Deployment | Single binary ~8MB | Cloud + self-host | Docker/Postgres | Docker/Postgres | Dependencies | None SQLite embedded | Python + vector DB | Postgres + Python | Postgres + Go | MCP-Native | ✅ 46 tools | ❌ Not MCP-native | ❌ Not MCP-native | ❌ Not MCP-native | Offline/Local | ✅ Fully local | Cloud-dependent | Docker needed | Docker needed | Encryption | AES-256-GCM ✅ | ❌ | ❌ | ❌ | Hybrid Search | BM25 + Dense + RRF | Vector only | Vector only | Vector + Graph | Entity Lifecycle | Decay + Promote + Archive | ❌ | ❌ | ❌ | Entity Graph | Link + Traverse | ❌ | ❌ | ✅ | Journal Audit Trail | ✅ Immutable | ❌ | ❌ | ❌ | State Management | ✅ Key-value + TTL | ❌ | ❌ | ❌ | MCP Tools | 46 | 5 | 8 | 0 | GitHub Stars | ~20 | ~55K | ~15K | ~3K | License | MIT | Apache 2.0 | Apache 2.0 | Apache 2.0 | Full comparison: Mimir vs Mem0 → /Perseus-Computing-LLC/mimir/blob/main/docs/comparison/mimir-vs-mem0.md vs Letta → /Perseus-Computing-LLC/mimir/blob/main/docs/comparison/mimir-vs-letta.md vs Zep → /Perseus-Computing-LLC/mimir/blob/main/docs/comparison/mimir-vs-zep.md Mimir handles production workloads on modest hardware: | Metric | Result | |---|---| 100K entity insert | 1.01s 98,732 entities/s | FTS5 recall 10 results | 0.022s | Decay tick 100K entities | 1.317s batched, transactional | Memory 100K entities | ~85MB RSS | DB file size 100K | ~45MB with FTS5 index | Run it yourself: cargo test stress 100k --release -- --ignored --nocapture Ready-to-use adapters that make Mimir the default memory backend for popular AI agent frameworks: | Framework | Integration | Type | |---|---|---| LangGraph | MimirStore BaseStore implementation CrewAI MimirMemoryTool AutoGen MimirMemory Memory implementationEach adapter: - Connects via MCP stdio subprocess persistent session - Maps the framework's memory interface to Mimir tools - Comes with a README quickstart 5 minutes to working - Has passing tests with mocked MCP transport Any MCP-compatible framework works with Mimir directly. See Awesome Mimir /Perseus-Computing-LLC/mimir/blob/main/awesome-mimir.md for the full list. | Tool | Description | |---|---| mimir remember | Store/update entity. Idempotent by category, key ; a content change snapshots the prior version into history. | mimir recall | Search with FTS5/dense/hybrid modes, filters, stemming expansion. | mimir recall layer | Recall from a specific biomimetic layer world, episodic, semantic . | mimir recall when | Proactive just-in-time recall: surface entities whose recall when triggers match. | mimir get entity | Fetch one entity by ID with full body json . | mimir as of | Bi-temporal time-travel: the version of a fact category + key that was live at a past instant. | mimir history | List every superseded version of a fact category + key , newest first — the full version trail companion to mimir as of . | mimir forget | Soft-delete archived=1 . | | Tool | Description | |---|---| mimir ask | RAG: recall context, query LLM, return grounded answer with sources. | mimir embed | Generate dense vectors via the bundled model, Ollama, or OpenAI-compatible endpoint. | mimir semantic search | Dense-only semantic search shortcut — find entities by meaning, ranked purely by embedding similarity no keyword fallback . | mimir context | Pre-formatted markdown block for session injection. | mimir ingest | Trigger connector syncs GitHub, file watcher . | mimir ingest file | Locally extract a document's text plaintext/markdown always; DOCX/PDF with the multimodal feature and store it as a recallable entity. | mimir extract | Local, deterministic, rule-based knowledge extraction facts / preferences / temporal events / episodes from text or a stored entity. Read-only. | | Tool | Description | |---|---| mimir link | Create typed relationship links between entities. | mimir unlink | Remove entity links. | mimir traverse | Walk entity link graph up to configurable depth. | | Tool | Description | |---|---| mimir journal | Append structured event with actor attribution. | mimir timeline | Query journal by time range with filters. | | Tool | Description | |---|---| mimir state set | Set key-value state with optional TTL. | mimir state get | Get state value. Returns null if expired. | mimir state delete | Delete state entry. | mimir state list | List state keys, optionally filtered by prefix. | | Tool | Description | |---|---| mimir decay | Recalculate Ebbinghaus decay scores batched 1000-entity transactions . | mimir prune | Bulk archive by category, decay threshold, or age. | mimir purge | Permanently delete archived entities + VACUUM. Destructive. | mimir cohere | Autonomous coherence grooming pass — promote, decay, link, archive. | mimir autocohere | Full atomic grooming: cohere → decay → compact in one pass supports dry-run . | mimir compact | Archive entities below decay threshold. | mimir reindex | Rebuild FTS5 search index from entities table. | | Tool | Description | |---|---| mimir score | Assign quality score 0.0-1.0 . | mimir conflicts | Detect conflicting entities via trigram similarity; opt-in resolve=true invalidates the lower-certainty side into history reversible, dry-run by default . | mimir correct | Structured correction capture for learning from errors. | mimir supersede | Mark a new fact as superseding an old one sets the old entity to deprecated . | | Tool | Description | |---|---| mimir vault export | Export entities to .md files with YAML frontmatter. | mimir vault import | Import from .md vault directory idempotent . | mimir federate | Copy entities between workspaces. | mimir share | Share one entity by category + key into another workspace, preserving content. | mimir workspace list | List all distinct entity categories. | | Tool | Description | |---|---| mimir stats | Full DB statistics across all tables. | mimir health | Server and DB health check. | mimir bench | Performance benchmark tracking. | mimir maintenance | DB maintenance: dedup, orphan detection, VACUUM, FTS5 reindex supports dry-run . | mimir synthesize | LLM session synthesis — extract lessons from transcripts. | mimir migrate | Migrate v0.1.x DB to current schema. | Server mimir serve --db /data/mimir.db mimir serve --web --port 8767 --encryption-key ~/.mimir/secret.key mimir serve --llm-endpoint http://localhost:11434/api/generate --llm-model llama3 mimir serve --transport sse --port 8787 --mcp-token my-secret-token Maintenance operate directly on DB, no server needed mimir stats --db /data/mimir.db mimir forget --db /data/mimir.db --category decision --key stale-choice --reason "superseded" mimir prune --db /data/mimir.db --category junk --min-decay 0.1 --dry-run mimir purge --db /data/mimir.db --dry-run mimir decay --db /data/mimir.db mimir reindex --db /data/mimir.db mimir vault-export --db /data/mimir.db --vault-dir ./export/ mimir vault-import --db /data/mimir.db --vault-dir ./export/ mimir obsidian-sync ~/obsidian-vault/Mimir/ one-shot export to an Obsidian vault mimir obsidian-sync ~/obsidian-vault/Mimir/ --watch continuous sync on every memory change Key management mimir keygen --key-file ~/.mimir/secret.key | Flag | Description | |---|---| --db | SQLite database path default: ~/.mimir/data/mimir.db | --web | Start web dashboard | --port | Dashboard port default: 8767 | --web-bind | Dashboard bind address default: 127.0.0.1 | --transport | MCP transport: stdio default , sse , or http | --mcp-token | Bearer token for SSE/HTTP transport auth | --encryption-key | AES-256-GCM key file path | --llm-endpoint | LLM API endpoint for mimir ask and embeddings | --llm-model | LLM model name default: llama3 | --llm-api-key | API key for LLM endpoints OpenAI, Azure, etc. | --embedding-endpoint | OpenAI-compatible embedding endpoint | --connectors-config | Path to connectors.yaml | Mimir is your AI agent's long-term memory — and it doubles as your second brain. Every entity your agent remembers exports to a plain Markdown note with YAML frontmatter, so your AI's memory becomes a navigable personal knowledge base inside the tools you already use: Obsidian, Logseq, or Notion. Export your entire memory to an Obsidian vault as linked Markdown notes mimir obsidian-sync ~/obsidian-vault/Mimir/ Keep it live — re-export automatically on every memory change mimir obsidian-sync ~/obsidian-vault/Mimir/ --watch Open the vault in Obsidian and you get a graph of your agent's knowledge. WikiLink backlinks. When one entity links to another via mimir link or a depends on / implements / references relationship , the exported note gets a Links section with WikiLink backlinks that resolve natively in Obsidian's graph view: --- id: cli-de8dfb8364b6 category: architecture key: api type: insight decay score: 0.5000 --- {"content":"axum service"} Links - cli-99756b494c7d|database depends on Links resolve by entity id notes are written as <id .md so they never break, and Obsidian shows the human-readable key as the link label. Open the graph view and your agent's architecture, decisions, and insights become a clickable knowledge map. --watch polls Mimir's cheap, deterministic state digest on an interval and re-exports only when memory actually changes. It naturally catches every mimir remember write with no filesystem-watcher dependency and no coupling to the server. Tune the interval with MIMIR SYNC INTERVAL SECS default: 2s .| Tool | How | |---|---| Obsidian | mimir obsidian-sync <vault — WikiLinks resolve in the graph view out of the box. | Logseq | Point obsidian-sync at your Logseq graph directory. Logseq reads the same WikiLink syntax and Markdown frontmatter. | Notion | Run mimir vault-export , then use Notion's Import → Markdown & CSV to pull the notes in. | Unlike cloud-only "second brain" tools, Mimir runs 100% local , is written in Rust , encrypts at rest with AES-256-GCM , and applies decay scoring so stale memories fade — your knowledge base stays yours and stays fresh. Bundled, in-process embeddings — a quantized all-MiniLM-L6-v2 model 384-dim is compiled into the binary, so dense/semantic search works with zero config and zero network : no Ollama, no API key, no model download. This is the default build bundled-embeddings feature . Auto-embed on write 271 — mimir remember embeds each new or content-changed entity synchronously as it is written, using the bundled model. Single-entity embedding is deterministic and LRU-cached, so it is cheap and adds no background tasks. Embedding failures are non-fatal logged to stderr ; the write always succeeds. Hybrid is the default recall mode 271 — mimir recall query=... with no mode flag automatically selects hybrid dense + keyword fused via RRF whenever embeddings exist, and transparently falls back to fts5 keyword search when none do. No manual mimir embed step, no flags to remember.— a one-tool shortcut for pure dense, meaning-based search no keyword fallback when you just want "find things like this". mimir semantic search query, limit Optional alternate embedder — to use Ollama or any OpenAI-compatible /v1/embeddings endpoint instead of the bundled model, set --llm-endpoint and --embedding-endpoint / --llm-api-key as needed . This is entirely optional; the bundled model is used by default.- Build a lean binary without bundled embeddings via cargo build --no-default-features — recall then defaults to keyword search unless a remote embedder is configured. FTS5 keyword search with LIKE fallback and Porter stemming expansion Dense vector search via cosine similarity on stored embeddings Reciprocal Rank Fusion RRF — combine keyword + vector results Query expansion — automatic stemming variants for broader recall Mimir models memory using three biomimetic layers, inspired by human memory pathways: World Core : Slow-decaying, global facts about the environment. Episodic Buffer : Fast-decaying, session-specific interaction history. Semantic Working : Medium-decaying, general knowledge and learned concepts. You can interact with these layers directly using the mimir recall layer tool or by specifying the layer parameter in mimir remember . Ebbinghaus decay — memories naturally fade unless retrieved refresh on access Layer promotion — buffer → working → core based on access frequency Automatic archival — stale entities archive; purge to permanently delete + VACUUM Always-on entities — pin critical memories for unconditional session injection — natural language Q&A over stored memories via any LLM Ollama, OpenAI, etc. mimir ask — generate and store dense vectors via Ollama or OpenAI-compatible mimir embed /v1/embeddings - Supports single-entity and batch-category embedding AES-256-GCM transparent encryption for entity body json - Opt-in via --encryption-key flag mimir keygen subcommand for key generation- FTS5 index stays plaintext for search - Built-in Axum HTTP server mimir serve --web --port 8767 - Dark-themed dashboard with search, entity table, vis.js graph, timeline - Default bind: 127.0.0.1 use --web-bind 0.0.0.0 to expose - Separate SQLite connection in WAL mode for concurrent reads GitHub issues connector — ingest issues/PRs by repo, rate-limit aware File watcher — scan directories for .md / .txt / .json files with content-hash dedup- YAML-based connector config via --connectors-config stdio default — zero-config, works with any MCP host SSE — Server-Sent Events for HTTP-based MCP clients HTTP — REST-style MCP endpoint Bearer token auth — for SSE/HTTP transports Mimir is the default memory backend for Perseus https://perseus.observer : mimir: enabled: true transport: "stdio" command: "mimir", "serve", "--db", "~/.mimir/data/mimir.db" timeout s: 30.0 merge strategy: "local first" fallback to local: true context categories: "decision", "architecture", "convention" context limit: 10 Mimir is built for government deployment from the ground up. | Capability | Status | |---|---| License | MIT — no copyleft, no GPL/AGPL | SBOM | | Air-gapped Encryption at rest Audit trail Supply chain For federal buyers: See docs/federal-buyers.md /Perseus-Computing-LLC/mimir/blob/main/docs/federal-buyers.md for procurement information, compliance status, and deployment models air-gapped, on-premises, classified environments . Perseus Computing LLC is a US-owned small business. SAM.gov registration in progress. NAICS: 541715, 541511, 541512. Mimir is a local-first MCP server — it runs entirely on your machine. No data collection. Mimir does not collect, transmit, or phone home any user data, usage statistics, or telemetry.- All data remains in your local SQLite database file. - All memory entities, journal entries, and state are stored locally in a SQLite database at the path you specify via --db . - Optional AES-256-GCM encryption at rest is available — when enabled, entity bodies are encrypted before storage. - No data is shared with Perseus Computing LLC or any third party. None. Mimir is fully air-gapped by default. No API calls, no cloud services, no external network requests.- The optional dense vector embeddings feature uses a locally-compiled model — no external embedding API is called. - You control retention: entities can be soft-deleted mimir forget , archived via decay/compact , or permanently purged mimir purge . - No automatic off-machine backup is performed. Email: privacy@perseus.observer mailto:privacy@perseus.observer GitHub: Perseus-Computing-LLC/mimir https://github.com/Perseus-Computing-LLC/mimir MIT — see LICENSE /Perseus-Computing-LLC/mimir/blob/main/LICENSE .