Weave logo

Weave

Self-hosted MCP access to connected SQLite workspaces with tables, relations, documents, workflows, and reversible changes.

Install Weave

Run the project’s own setup commands, in order.

Setup steps from the project

weave installs with its own commands rather than a config we can generate, so these are copied as-is. Run them in order in a terminal. Source: github.com

  1. 1. Start the server
    node bin/weave.js mcp --data <file.db>
    
  2. 2. Add to Claude Code
    claude mcp add weave -- node /path/to/weave/bin/weave.js mcp --data /path/to/workspace.db
    
  3. 3. Install
    git clone https://github.com/grunion-ai/weave
    
  4. 4. Run
    cd weave
    
  5. 5. Start the server
    docker compose up -d
    
  6. 6. Start the server
    docker run -d --name weave -p 127.0.0.1:4400:4400 -v weave-data:/data ghcr.io/grunion-ai/weave:latest
    

Overview

weave MCP server connects AI agents to self-hosted workspaces stored in SQLite files. Agents can build and query connected tables, update rows, manage relations, work with markdown documents, and reverse changes through MCP tools. Use it when an agent needs structured work data alongside documents and workflows without relying on a hosted SaaS database. It supports local stdio connections and hosted HTTP endpoints with sign-in.

Use cases

  • Build a connected project workspace from a JSON specification
  • Query task tables and related records from an AI agent
  • Update rows, links, comments, and workflow states through MCP
  • Restore document revisions or undo workspace changes
  • Serve a local SQLite work database to multiple agent clients

Key features

  • SQLite-backed workspaces
  • Connected tables, relations, formulas, and rollups
  • Markdown documents with version history
  • MCP tools for building, querying, updating, and undoing changes
  • Local stdio and hosted HTTP connections
  • Role-filtered tool listings

Security & health

What it can reach and how reliably it runs — from our own checks, not the maintainer’s claims.

ActiveLast checked Today

Recent health check succeeded.

7/7 checks healthy over the last 25d

Supply-chain signal

No high-severity advisories surfaced by our automated scan.

Critical 0High 0Medium 0Low 0

Scanned 17d ago via OSV.dev

Adoption & maintenance

Factual signals from GitHub, npm, and our automated checks — not a rating.

GitHub stars15Stargazers on the source repository.
Last commitTodayMost recent push to the default branch.
Directory activity1 viewsConfig copies, upvotes, and views on AllMCPs.

How Weave works

What weave MCP server does

weave MCP server gives AI agents access to a self-hosted work database whose workspace is stored in a SQLite file. A workspace contains spaces, tables, rows, relations, workflow states, formulas, rollups, automations, and markdown documents attached to rows. The same data can also be used through weave's web UI, REST API, and CLI.

The MCP interface is intended for agents that need to create or operate on structured workspaces. Rows are called entities by the API and can be referenced with identifiers such as Task#12. Documents can be addressed as Markdown, HTML, or PDF URLs, while document revisions can be browsed and restored.

How it works

A local process runs the MCP transport against a selected SQLite database. The standard command is node bin/weave.js mcp --data <file.db>. An MCP client launches that process over stdio and sends tool calls to the workspace represented by the database file.

The server exposes 17 tools by default, including the tools used by a workspace build, weave_build, and weave_call. weave_call can run the other 48 tools. Passing --tools all, or setting WEAVE_MCP_TOOLS=all, exposes all 65 tools. The complete set includes workspace operations, undo, document revision handling, and weave_vocabulary, which returns allowed values for fields.

Hosted deployments use an HTTP endpoint instead. Each workspace has an endpoint under /w/<workspace>/mcp; an OpenID Connect sign-in provider can authenticate the connection. The available tools are filtered according to the connected account's role.

Setup and configuration

weave requires Node.js 22.16 or newer. Clone the repository and run the MCP command directly; the project does not require an npm installation or build step. Point --data to the SQLite file that contains the target workspace.

Example local configuration:

config.json
{
  "mcpServers": {
    "weave": {
      "command": "node",
      "args": [
        "/path/to/weave/bin/weave.js",
        "mcp",
        "--data",
        "/path/to/my-workspace.db"
      ]
    }
  }
}

Claude Code can register the same connection with claude mcp add. For a hosted instance, use the HTTP endpoint with --transport http; the first call opens the provider's sign-in flow in a browser.

Tools and capabilities

  • Build a workspace from a specification with weave_build or the weave build command.
  • Read and modify tables, rows, relations, workflows, formulas, and rollups through the MCP tool set.
  • Run additional operations through weave_call when the full tool set is not listed individually.
  • Reverse row edits, creations, deletions, links, and comments with weave_undo.
  • Inspect and restore document revisions with weave_doc_revisions and weave_doc_restore.
  • Query allowed field values with weave_vocabulary instead of guessing them.

Limitations and notes

Each local workspace is represented by one SQLite file and is served by a Node process. The local setup does not require an account and is designed for self-hosting on a laptop or server. Hosted connections require account linking and provider sign-in.

The default tool listing is intentionally smaller than the complete interface. Clients that need tools outside the default 17 must start the server with --tools all or set WEAVE_MCP_TOOLS=all. The HTTP endpoint also limits listed tools based on the connected role.

The project is licensed under MIT and has no runtime dependencies beyond Node's built-in SQLite support and vendored, pinned third-party code.

Reviews

No reviews yet — be the first to share how it worked for you.

Frequently Asked Questions about Weave

weave is an MCP server that connects AI agents to self-hosted SQLite workspaces containing tables, relations, workflows, and Markdown documents. Its main tools build workspaces, query and update rows, call additional operations, return allowed field values, undo changes, and restore document revisions through local stdio or hosted HTTP connections.

Clone the repository and run `node bin/weave.js mcp --data <file.db>`; no npm installation or build step is required. Configure your MCP client with that command and the path to the workspace SQLite file. Node.js 22.16 or newer is required.

Start weave with `--tools all` or set `WEAVE_MCP_TOOLS=all`. The default listing contains 17 tools, while the full interface lists 65, including the operations available through `weave_call`.

Yes, `weave_undo` reverses row edits, creations, deletions, links, and comments. The same undo capability is also available through weave's CLI and REST API.

Yes, weave documents Cursor as a supported stdio MCP client. Add a server entry that runs the weave command with the `mcp` and `--data` arguments pointing to the target SQLite workspace.

Related MCP servers

Add a badge to your READMEShow it’s listed — live SVG, Markdown or HTML.

AllMCPs Directory Badge

Full Badge Customizer

Showcase your server listing on GitHub or your project documentation. Embed this dynamic SVG badge to highlight official listing status and live engagement.

Badge Style:
Live Dynamic SVG PreviewWeave AllMCPs Directory Badge
Markdown (GitHub README)
[![AllMCPs](https://allmcps.com/api/badge/weave?style=directory)](https://allmcps.com/mcp/weave)
HTML Embed
<a href="https://allmcps.com/mcp/weave"><img src="https://allmcps.com/api/badge/weave?style=directory" alt="Weave on AllMCPs" /></a>