Synapse MCP Server
A Model Context Protocol (MCP) server that enables AI agent access to Synapse entities (Datasets, Projects, Folders, Files, Tables, and more).
Power your agent
Instal dari penerbit. Bukan endpoint PlugMCP.
URL remote atau perintah di mesin Anda? Perbedaan kedua cara instal
Jalankan di PlugMCP
A new server you create. It is not this listing's official endpoint. Use the publisher install when you want their URL and their credentials.
BukaDari penerbit
A Model Context Protocol (MCP) server that enables AI agent access to Synapse entities (Datasets, Projects, Folders, Files, Tables, and more).
A Model Context Protocol (MCP) server that enables AI agent access to Synapse. Deployed into AWS ECS via https://github.com/Sage-Bionetworks-IT/synapse-mcp-infra
You (your AI agent) can:
The catalog includes both read tools and write/destructive tools (create/update/delete) across the entity, schema, team, evaluation, submission, organization, and curation domains. The server never uploads or downloads file content — File entities are created only via an external URL or an existing file handle, never from local file bytes. The single exception is create_record_based_curation_task, which uploads a header-only CSV template that the server generates from a registered JSON schema's property names, since a RecordSet must be backed by a CSV.
Read and write tools are dispatched through separate proxies — call_read_tool and call_write_tool — so a client that gates permissions by tool name can allow reads while withholding writes to run the server read-only.
Resources provide ready-to-present context that clients can pull without extra parameters. When you need to search or compute derived results, prefer tools instead.
Important: When using this MCP server with external AI services (such as Claude, ChatGPT, or other cloud-based models), please be aware that:
Recommended Safe Usage:
You are responsible for ensuring your usage complies with the Synapse Terms of Service.
The Synapse MCP server can be used as a remote hosted server (recommended) or installed locally from source. Choose the approach that fits your needs.
The hosted server is available at:
Authentication uses OAuth2 -- your MCP client will open a browser window for you to log in to Synapse. No API keys or tokens to manage.
Below are setup instructions for popular AI clients. If your client is not listed, use the generic JSON config.
Most MCP-compatible clients accept a JSON configuration block. Add the following to your client's MCP config file:
{
"mcpServers": {
"synapse": {
"url": "https://mcp.synapse.org/mcp",
"type": "http"
}
}
}
Go to Settings > Connectors > Add custom connector and enter the URL https://mcp.synapse.org/mcp.
claude mcp add --transport http synapse -- https://mcp.synapse.org/mcp
VS Code's MCP client does not yet fully support OAuth Dynamic Client Registration (DCR). To connect to the remote server, follow these steps:
Step 1: Register a client
Run this command once to register an OAuth client with the MCP server:
curl -X POST https://mcp.synapse.org/register \
-H "Content-Type: application/json" \
-d '{
"client_name": "vscode-synapse",
"redirect_uris": ["http://127.0.0.1"],
"grant_types": ["authorization_code", "refresh_token"],
"response_types": ["code"],
"token_endpoint_auth_method": "none"
}'
Save the client_id from the response (e.g., c3dfaf80-126c-4f46-80ab-114747fcc3b3).
Step 2: Configure VS Code
Create or edit .vscode/mcp.json in your workspace:
{
"servers": {
"synapse": {
"url": "https://mcp.synapse.org/mcp",
"type": "http"
}
}
}
Step 3: Complete the OAuth flow
When you start the server, VS Code will open a browser to an authorization URL. Replace the client_id value in the URL with your registered client_id from Step 1, then press Enter to continue the Synapse login flow.
For example, change client_id=100441 to client_id=YOUR_CLIENT_ID in the browser address bar.
Alternatively, you can use the Local Server setup with a Personal Access Token, which does not require OAuth.
Add to Cursor Settings > MCP > + Add new global MCP server, or add to your project's .cursor/mcp.json:
{
"mcpServers": {
"synapse": {
"url": "https://mcp.synapse.org/mcp",
"type": "http"
}
}
}
Run the server locally for development, self-hosting, or offline use. The local server uses stdio transport by default, which is what most MCP clients expect for command-based servers.
Note:
synapse-mcpis not currently published on PyPI. You must install from source.
git clone https://github.com/Sage-Bionetworks/synapse-mcp.git
cd synapse-mcp
python -m venv .venv
source .venv/bin/activate # On Windows: .venv\Scripts\activate
pip install -e .
After installation, the synapse-mcp command is available in your virtual environment.
For local use, authenticate with a Synapse Personal Access Token (PAT) by setting the SYNAPSE_PAT environment variable:
export SYNAPSE_PAT="your_synapse_pat_here"
To create a PAT, visit your Synapse Personal Access Tokens page.
Important: The
synapse-mcpcommand must be on your PATH. If you installed in a virtual environment, either activate it first or use the full path to the binary (e.g.,/path/to/.venv/bin/synapse-mcp).
Generic JSON config (stdio):
{
"mcpServers": {
"synapse": {
"command": "/path/to/.venv/bin/synapse-mcp",
"env": {
"SYNAPSE_PAT": "your_synapse_pat_here"
}
}
}
}
Claude Code (local):
claude mcp add synapse -e SYNAPSE_PAT=your_synapse_pat_here -- /path/to/.venv/bin/synapse-mcp
VS Code / GitHub Copilot (local):
In .vscode/mcp.json:
{
"servers": {
"synapse": {
"command": "/path/to/.venv/bin/synapse-mcp",
"env": {
"SYNAPSE_PAT": "your_synapse_pat_here"
}
}
}
}
Cursor (local):
{
"mcpServers": {
"synapse": {
"command": "/path/to/.venv/bin/synapse-mcp",
"env": {
"SYNAPSE_PAT": "your_synapse_pat_here"
}
}
}
}
You can configure which Synapse platform instance to connect to by setting the SYNAPSE_ENV environment variable:
prod (default) -- Production instance at synapse.orgstaging -- Staging instance at staging.synapse.orgdev -- Development instance at dev.synapse.orgIf not set, the server defaults to prod.
For contributor/development setup details, see DEVELOPMENT.md.
To keep the LLM's context small, the server does not expose the full tool catalog on every call. The default view is just two always-visible tools — get_entity and search_synapse — plus a synthetic search_tools / call_tool pair. To reach any other tool, the LLM issues a natural-language query to search_tools (a BM25-ranked search over tool names, descriptions, synonyms, and siblings) and then invokes the chosen tool via call_tool. See doc/tool-authoring.md for the conventions each tool follows.
See usage examples
Contributions are welcome! Please see our Development Guide for instructions on setting up a development environment, running tests, and more.
MIT
For issues, please file an issue. For other contact, see https://sagebionetworks.org/contact.
A Model Context Protocol (MCP) server that enables AI agent access to Synapse. Deployed into AWS ECS via https://github.com/Sage-Bionetworks-IT/synapse-mcp-infra
Tool
Belum ada daftar tool. Server remote mendapatkannya setelah pemeriksaan kesehatan pertama.
FAQ
Salin konfigurasi stdio untuk Cursor, Claude Desktop, Claude Code, atau VS Code di halaman ini. Konfigurasi ini memakai instalasi dari penerbit, bukan proxy PlugMCP.
Tidak ada variabel lingkungan wajib yang tercantum di catatan ini.
Synapse berjalan secara lokal melalui stdio.
Terakhir diperiksa 2026-10-28. A Model Context Protocol (MCP) server that enables AI agent access to Synapse. Deployed into AWS ECS via https://github.com/Sage-Bionetworks-IT/synapse-mcp-infra