MCP Project Bridge
Enables multi-project workspaces to share structured notes, API contracts, and handoff messages via a local SQLite database, with versioning and read tracking.
README
MCP Project Bridge
MCP Project Bridge is a stdio MCP server for sharing project messages through a local SQLite database. It is designed for multi-project workspaces where one project needs to leave structured notes, API contracts, implementation status, or handoff messages for another project.
The server writes only to its own central database. It does not create files inside the projects being bridged.
How It Works
Each project is represented by a project key. You can provide your own key, or derive a stable key from an absolute project path with derive_project_key.
Messages are stored in a central inbox model:
- A direct message has a sender project key, a target project key, and a
docKey. - A broadcast message has a sender project key and a
docKey, with no target project key. - Reusing the same sender, target, and
docKeycreates a new version of the same message. - Passing
messageIdupdates that message, but only the sender project can update it. - Read state is tracked per viewer project and per message version.
read_unread_messagesmarks only the returned latest versions as read.- A later message version becomes unread again for projects that have not read that version.
Broadcast messages are hidden from normal inbox reads unless withBroadcast: true is passed.
Local Setup
Requirements:
- Node.js 20 or newer
- Yarn 1.x
Install and build:
yarn install
yarn build
Configure your MCP client to run the built stdio server:
{
"mcpServers": {
"project-bridge": {
"command": "node",
"args": ["D:\\mcp\\mcp-project-bridge\\dist\\index.js"]
}
}
}
Use an absolute path to dist/index.js. On Windows JSON paths need escaped backslashes.
Database Location
By default, the database is stored in the user data directory:
- Windows:
%APPDATA%\\mcp-project-bridge\\bridge.sqlite - Linux and WSL:
$XDG_DATA_HOME/mcp-project-bridge/bridge.sqlite, or~/.local/share/mcp-project-bridge/bridge.sqlite
Set MCP_PROJECT_BRIDGE_DB to choose a database file:
{
"mcpServers": {
"project-bridge": {
"command": "node",
"args": ["D:\\mcp\\mcp-project-bridge\\dist\\index.js"],
"env": {
"MCP_PROJECT_BRIDGE_DB": "D:\\mcp\\project-bridge-data\\bridge.sqlite"
}
}
}
}
Docker Image
Build the image:
docker build -t mcp-project-bridge:latest .
Run it as a stdio MCP server from an MCP client:
{
"mcpServers": {
"project-bridge": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"-v",
"mcp-project-bridge-data:/data",
"mcp-project-bridge:latest"
]
}
}
}
The image sets MCP_PROJECT_BRIDGE_DB=/data/bridge.sqlite, so mount /data to keep messages across container restarts. The server uses stdio and does not expose an HTTP port.
Project Keys
derive_project_key normalizes Windows drive paths into WSL-style paths before replacing path separators with underscores.
Examples:
D:\mcp\api -> /mnt/d/mcp/api -> _mnt_d_mcp_api
C:/work\frontend -> /mnt/c/work/frontend -> _mnt_c_work_frontend
/home/me/web -> _home_me_web
Project keys are just message identifiers. They do not need to be registered before use.
Tools
derive_project_key
Derive a stable project key from an absolute project path without writing to storage.
{
"path": "D:\\workspace\\backend"
}
upsert_message
Create or update a direct or broadcast message.
{
"currentProjectKey": "_mnt_d_workspace_backend",
"targetProjectKey": "_mnt_d_workspace_frontend",
"docKey": "users-api",
"title": "Users API",
"content": "GET /users\nPOST /users",
"format": "markdown",
"tags": ["api", "users"]
}
Omit targetProjectKey, pass null, or pass an empty string to create a broadcast message:
{
"currentProjectKey": "_mnt_d_workspace_backend",
"docKey": "release-note",
"title": "Backend Release",
"content": "Backend release is ready",
"tags": ["release"]
}
Supported formats are markdown, text, and json. The default format is markdown.
read_unread_messages
Read unread inbox messages for the current project and mark the returned latest versions as read.
{
"currentProjectKey": "_mnt_d_workspace_frontend",
"withBroadcast": true,
"limit": 20
}
list_messages
List latest inbox message summaries without changing read state.
{
"currentProjectKey": "_mnt_d_workspace_frontend",
"withBroadcast": true,
"query": "users",
"tags": ["api"],
"limit": 20
}
get_message_history
Read version history for a message without changing read state.
{
"currentProjectKey": "_mnt_d_workspace_frontend",
"messageId": 1,
"withBroadcast": true,
"limit": 10
}
Direct message history is visible to the sender and target project. Broadcast history is visible to the sender, and to other projects only when withBroadcast: true is passed.
Typical Workflow
- Derive or choose the project key for each workspace.
- Have the sender call
upsert_messagewith a stabledocKey. - Have the receiver call
read_unread_messages. - Use
list_messageswhen you need search or status without marking anything read. - Use
get_message_historywhen you need prior versions of a message.
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
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.
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.