A local-first, LLM-agnostic memory layer for AI assistants — runmemento.com
Every AI session starts the same way: re-explaining your preferences, your project's conventions, the decisions you made last week, the dead-ends to avoid. Each tool solves this in its own siloed way (CLAUDE.md, .cursorrules, copilot-instructions.md, ChatGPT Memory) — but you, the human, are the one constant. Your memory shouldn't fragment across vendors.
Memento is one place where that memory lives. It runs an MCP server over a local SQLite file, so any MCP-capable AI assistant — Claude Desktop, Claude Code, Cursor, GitHub Copilot, Cline, OpenCode, Aider, a research bot, a custom agent — can read and write durable, structured memory about you, your work, and your decisions. Local-first, no outbound network calls by default, no vendor lock-in.
npx @psraghuveer/memento initOn a TTY, init walks you through four one-keystroke setup questions: your preferred display name, whether to install the bundled skill into ~/.claude/skills/, whether to seed your store with a starter pack, and whether to auto-install the persona snippet into every detected file-based AI client's user-scope custom-instructions file (~/.claude/CLAUDE.md for Claude Code, ~/.config/opencode/AGENTS.md for OpenCode, ~/Documents/Cline/Rules/memento.md for Cline). Then prints copy-paste MCP snippets for every supported client, plus per-client paste instructions for UI-only clients (Cowork, Claude Desktop, Claude Chat, Cursor User Rules). Idempotent — re-run any time. Pass --no-prompt to suppress the interactive flow (CI, scripts).
For a persistent memento on your $PATH instead of npx:
npm install -g @psraghuveer/mementoRequires Node.js >= 22.11.
Three steps from zero to a working memory layer:
npx @psraghuveer/memento initCreates the SQLite database under the XDG default, runs migrations, walks you through four interactive setup questions on a TTY (preferred name, install the bundled skill, seed a starter pack, auto-install the persona snippet into detected file-based clients), and prints copy-paste MCP snippets for every supported client.
init prints either a one-line subcommand (where the client ships one) or a JSON snippet to merge into the client's MCP config. Pick the one for your client, paste it, then restart the client so it loads the new MCP server.
Per-client walkthrough: docs/guides/mcp-client-setup.md.
Memento ships three teaching surfaces — an MCP instructions spine on the wire (ADR-0026), a bundled skill for skill-capable clients, and the persona snippet. Of the three, only the persona snippet is guaranteed to reach the assistant's system prompt on every message — the MCP spec leaves instructions optional and current client implementations vary in whether they surface it; the skill is intent-triggered so it doesn't fire on neutral first messages.
If you said Y to the persona auto-install prompt in step 1, this is already done for every file-based AI client detected on your machine — the snippet was written to ~/.claude/CLAUDE.md, ~/.config/opencode/AGENTS.md, and/or ~/Documents/Cline/Rules/memento.md as applicable. For UI-only clients (Cowork, Claude Desktop, Claude Chat at claude.ai, Cursor User Rules) and any client you have where the auto-installer didn't fire, copy the persona snippet from teach-your-assistant.md and paste it into the client's custom-instructions / system-prompt slot (the field's name varies — "Custom Instructions" in the client's settings UI, etc.).
Verify the wiring end-to-end:
npx @psraghuveer/memento verify-setupSpawns the server as a subprocess and round-trips a write/search/cleanup through the MCP transport to confirm your assistant can actually use Memento.
Four pillars: Local. Typed. Audited. Yours.
- Local — single SQLite file under your home directory. No cloud, no telemetry, no outbound network calls by default. Fully offline. Built-in regex scrubber strips secrets before persistence; patterns are user-configurable.
- Typed — five memory kinds (
fact,preference,decision,todo,snippet) with kind-specific fields. The assistant can reason about what's still true, not just retrieve text blobs. - Audited — every write produces an event in an append-only log. A post-write hook surfaces conflicts for triage. Memories decay if you don't confirm them. Answer "why is this here?" and "when did it change?" at any time.
- Yours — every retrieval weight, decay half-life, and scrubber rule is configurable. JSONL
export/importfor backup, migration, or carrying memory between machines. Apache-2.0. Leave any time.
Plus: MCP-native (standard MCP server over stdio — Claude Desktop, Claude Code, Cursor, Copilot, Cline, OpenCode, Aider, custom agents — works without code changes), LLM-agnostic (no model dependencies; works with whatever LLM your client talks to), vector retrieval (optional, runs on a local embeddings model — no cloud calls), and packs — curated YAML bundles of memories you can install in one step. Author your own with memento pack create, share as a file, an HTTPS URL, or a community contribution. Four bundled packs ship out of the box (engineering-simplicity, pragmatic-programmer, twelve-factor-app, google-sre); install with memento pack install <id>. Full guide: packs.md.
- runmemento.com — overview, comparisons, FAQ
- README on GitHub — full quickstart, install, MCP client setup
- ARCHITECTURE.md — internal design, data model, retrieval pipeline
- CLI reference — every command and flag
- MCP tools reference — every tool exposed to clients
- Architecture Decision Records — decision history with rationale
The package is @psraghuveer/memento; the binary it installs is memento. The CLI is a structural mirror of @psraghuveer/memento-server — both are adapters over the @psraghuveer/memento-core command registry; see ADR-0003. Lifecycle commands (init, serve, dashboard, context, doctor, verify-setup, status, ping, backup, export, import, pack, store migrate, completions, explain, skill-path, uninstall) sit alongside a generic projection of the registry surface (memento <namespace> <verb>); the full surface is documented in the CLI reference.
Apache-2.0 licensed. Contributions welcome — see CONTRIBUTING.md.