moneta-notes-mcp
MCP server exposing fast full-text and semantic search over an Obsidian vault, plus safe structured note read and write access for Claude.
README
Moneta Notes (mnotes)
A personal notes system built around an Obsidian vault, providing fast full-text and semantic search plus safe, structured note read/write access for Claude via MCP.
Three components, one shared core library:
- Indexing Subsystem — a background daemon that keeps a SQLite index (FTS5 + sqlite-vec) in sync with the vault on disk.
- CLI Tools — terminal commands for manual/debug search and note access, used directly and from NeoVim.
- MCP Server — exposes search, read, and write capabilities to Claude Code / Claude Desktop.
The CLI and MCP server are thin wrappers around one shared core library. Neither implements
search, hashing, or file I/O logic itself — that guarantees the two surfaces can never drift out of
sync.
Runs on macOS only (Apple Silicon), on either a personal or work machine, never both at once. Each machine gets its own vault, its own index, and its own config.
This file is a maintainer-facing overview — architecture, tech stack, dev conventions. Detailed,
binding functional specs live in docs/specs/ and are the actual source of truth for what this
project does; if anything here conflicts with a spec, the spec wins.
Name
Moneta was an epithet of the Roman goddess Juno — Juno Moneta, "Juno who warns/reminds," from the Latin verb monere (to remind, to warn, to advise). Her temple on the Capitoline Hill also housed Rome's mint, which is why moneta is the root of "money" and "mint" in English and the Romance languages. The reminding/advising sense is the one that gave this project its name.
Documentation
| Guide | Covers |
|---|---|
| Installation | Prerequisites, running scripts/install.sh, what it creates, verifying the install |
| Configuration | Every config.toml option: what it does, its default |
| Usage | The full mnotes CLI command reference plus the MCP tool surface |
| Process Management | Starting/stopping/restarting the indexing daemon, log locations, troubleshooting |
| Uninstallation | Running scripts/uninstall.sh, what it removes, what it leaves alone |
Prerequisites
Runs on macOS only (Apple Silicon). Required: Node.js, pnpm. Strongly recommended: ripgrep, fswatch, Xcode Command Line Tools. Optional: the Claude Code CLI, for automatic MCP registration. See Installation for what each is for and what happens if it's missing.
License & Contributing
Licensed under Apache-2.0. Contributions are welcome — see CONTRIBUTING.md for signed-commit and DCO sign-off requirements, and Architecture § Development for running tests/lint and what the Husky hooks enforce.
Specs
| Spec | Covers |
|---|---|
| S001 — Data Model | SQLite schema: notes, FTS5, chunks/vectors, tags, links, index queue, migrations |
| S002 — Search | Fulltext/semantic/hybrid ranking, RRF merge, chunk collapse |
| S003 — Notes | Note CRUD, hashing, frontmatter/metadata, the id field, rename (incl. link cascade), backlinks |
| S004 — Grep & Tags | Ripgrep wrapper, tag extraction and listing |
| S005 — Indexing Daemon | fswatch loop, embedding pipeline, queue/retry, daemon↔CLI IPC |
| S006 — CLI | mnotes command surface and output formats |
| S007 — MCP Server | Tool schemas, transport, error mapping |
| S008 — Logging | Logger, rotation, audit trail |
| S009 — Config & Install | config.toml schema, install/uninstall, launchd |
| S010 — Shared Utilities | Title↔path conversion, line counting, code-region stripping |
| S011 — Links | Wikilink extraction, backlinks, rename link-cascade, broken-link detection |
Specs are numbered (S001, S002, ...) and named with a short kebab-case topic. A spec may amend an
earlier one (noted at the top of the file) as later design work surfaces a gap — e.g. S004 amends
S001's tags table collation, S005 amends S001 with the index_queue table.
Not yet speced: MCP prompts (weekly review automation, note triage, stale/orphan note detection, weekly note scaffolding — candidates named below, design deferred to a future spec once the core tool set is built and in use).
Architecture
Three components share one core/ library — cli/ and mcp/ are thin wrappers that do their own
argument parsing / protocol plumbing and then call the same functions underneath, so the two surfaces
can never drift out of sync. Plain JS, no build step.
See Architecture for the full directory layout and tech stack table.
Note Versioning
Version-controlling the notes vault (as distinct from this code repository) is out of scope for
mnotes itself, but strongly recommended — note_write/note_edit's write-safety guards are
preventive, not a substitute for being able to recover a prior version of a note. See
Note Versioning for a recommended, low-effort setup.
Status
Fully speced (see docs/specs/): data model, search, notes, grep/tags, indexing daemon, CLI, MCP
server, logging, config & install.
Not yet speced: MCP prompts (weekly review automation, note triage, stale note detection, weekly note scaffolding, orphan note identification) — candidates identified, design deferred until the core tool set above is built and in use.
Not yet implemented: everything — specs are complete, implementation hasn't started.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.