filamental-mcp

A localModel Context Protocolserver that connects AI assistants (Claude Desktop, Claude Code, etc.) directly to yourFilamentalknowledge graph.

- Filamentalinstalled and at least one vault opened (this initialises the SQLite index)
- Node.js 22 or later

{ "mcpServers": { "filamental": { "command": "node", "args": [ "--no-warnings", "/absolute/path/to/node_modules/filamental-mcp/dist/index.js" ] } } }

No--vaultargument needed. The server reads the active vault from Filamental automatically and reconnects when you switch worlds. To pin to a specific vault (e.g. for testing), pass--vault <absolute-path>explicitly.

{ "mcpServers": { "filamental": { "command": "npx", "args": [ "filamental-mcp", "--vault", "/absolute/path/to/your/vault" ] } } }

A connector'sdirectionis one ofnone,outgoing,incomingorbidirectional. Nothing else is accepted, and an unrecognised value is rejected rather than stored.

Onwrites(create_edge,create_node,update_node) direction is stated relative tosource→target:outgoingdraws the arrowhead at the target,incomingdraws it back at the source,bidirectionaldraws both,noneis a plain line.

Onreads(get_connections) direction is instead stated relative to the node you asked about, because that is what the user sees on screen:

This distinction matters. Which end of a connector is stored assourceis decided by whichever end the user happened to drag from when drawing it, and is invisible on the graph — an undirected connector looks identical either way round. So the same arrow reads asoutgoingfrom one end andincomingfrom the other, andget_connectionsflips it for you. Filtering withdirection: "outgoing"gives you edges whose arrow pointsaway fromthe node you asked about, never edges that merely happen to be stored with it assource. Abidirectionaledge matches bothoutgoingandincomingfilters, since it genuinely points both ways;undirectedmatches only edges with no arrow at all.

Usestored_ononly if you are editing the underlying Markdown file directly. It says nothing about what the arrow does.

filamental-mcp --vault <path> Use vault at <path> filamental-mcp --vault <path> --db <path> Override the SQLite database path (for testing)

Filamental stores all node data as Markdown files with YAML frontmatter inside your vault folder. It also maintains a SQLite index (stored in your OS app-config directory, not inside the vault) for fast full-text search and graph traversal.

This server opens that SQLite index read-write. Read tools query it directly. Write tools update both the Markdown file on disk and the SQLite index so the Filamental app sees changes immediately on next load.

The server keeps working even on a schema mismatch (the DB is a disposable index, so most read/write operations tolerate drift). If a tool call fails for an unrelated reason while versions are mismatched, the error message is annotated with which side to update.

- The pre-built binary (better-sqlite3) is Windows x64 only. Other platforms require building from source.
- Auto-config via Filamental Settings has been tested on Windows. macOS path resolution is included but untested.

Adaptive MCP memory system for AI applications. Learns which retrieval strategies work for your data, scores results using cognitive science models, builds a knowledge graph automatically, and validates every parameter change against real query history before adopting it. Patent pending.

Highly efficient context management for agentic AI: MCP code search, evidence packs, graph context, and memory for large projects.

Self-hosted Rust-based MCP server for AI agent memory — persistent, queryable memory with hybrid search, knowledge graphs, built-in embeddings, and 14 core tools (expandable to 86+ with profile-based tiering).

Authenticated MCP and agent gateway for Forge Cascade private AI memory, provenance, graph search, and capsule lineage.

Source-grounded MCP, A2A, and x402 metadata for verified knowledge commerce.

Rust MCP server that gives AI coding agents architectural awareness — persistent knowledge graph, blast radius analysis, co-change detection via git, and local semantic search. No API keys, runs entirely on-premise.

Self-improving MCP memory server for AI agents. One command, no cloud, no config. Hybrid search, feedback loop quality system, dashboard UI, auto-linking knowledge graph. Gets better the more you use it.

Persistent memory for AI agents using a semantic knowledge graph. Store, retrieve, and connect memories with semantic search — so your AI remembers context across sessions.

Self-hosted memory and knowledge-graph MCP server for AI agents. TypeScript on Postgres 16 + pgvector, 11 brain_* tools. Runs keyless: full-text and semantic search, ingestion, content-hash dedup, provenance via brain_why, and the knowledge graph all work with no API keys via local Ollama. Apache-2.0.

Local-first MCP server that turns Obsidian and Markdown knowledge bases into an agent-readable graph