Synapse.org MCP Server

The Synapse MCP server gives AI agents read access to entities on Synapse.org, the collaborative scientific data platform operated by the nonprofit Sage Bionetworks. Through the server, agents can fetch core metadata, custom annotations, and provenance (activity) history for projects, folders, files, datasets, and tables, list the children of container entities, and run keyword searches with optional name, type, and parent filters.

The canonical hosted endpoint runs at https://mcp.synapse.org/mcp on Sage Bionetworks infrastructure. Authentication uses OAuth2 by default with a browser-based login flow, or a Synapse Personal Access Token for environments where a browser is not available. Access permissions follow the user's existing Synapse account, so the agent only sees entities the user is entitled to view.

The server is useful for researchers and informaticists working with public biomedical datasets hosted on Synapse, including projects from the AD Knowledge Portal, NF Data Portal, dbGaP-linked studies, and other community resources. The source implementation lives at susheel/synapse-mcp and also exposes a synapse://feeds/blog resource for the latest Sage Bionetworks publications.

Tools

Hosted endpoint

The server is provider-hosted at https://mcp.synapse.org/mcp. You authenticate with your Synapse account via OAuth2 in the browser, or with a Personal Access Token if a browser flow is not available.

Prerequisites:

• A free Synapse account at synapse.org
• For PAT-based auth: a Personal Access Token from your Synapse account settings

Claude Code

claude mcp add --transport http synapse -- https://mcp.synapse.org/mcp

Claude Desktop

Open Settings, Connectors, Add custom connector, then enter:

https://mcp.synapse.org/mcp

Generic MCP client (HTTP transport)

{
  "mcpServers": {
    "synapse": {
      "transport": "http",
      "url": "https://mcp.synapse.org/mcp"
    }
  }
}

Self-host

The source lives at susheel/synapse-mcp. See DEVELOPMENT.md in the repo for local install, dependency, and environment variable setup if you want to run your own instance.

Compliance note

The README warns that data sent to external AI providers may be stored, logged, or used for model training. Review Synapse's Terms of Service before using the server with sensitive or controlled-access data, and prefer self-hosted deployments for restricted datasets.

• Pull metadata and annotations for a specific Synapse entity ID inside an AI chat without leaving the assistant
• Audit the provenance graph of a derived dataset to understand which inputs and activities produced a given file or table version
• Browse the structure of a Synapse project or folder by listing its children, then drilling into specific files
• Discover relevant datasets across Synapse by keyword search with filters on entity type or parent project
• Stay current with Sage Bionetworks publications via the built-in synapse://feeds/blog RSS resource

• "Get the metadata and annotations for Synapse entity syn12345678."
• "List the child folders and files under project syn4939902."
• "Search Synapse for datasets related to Alzheimer's disease RNA-seq and show me the top results."
• "Show the provenance for syn22000001 version 3, including which inputs and activities produced it."
• "What are the latest posts from the Sage Bionetworks blog?"

• Provider-hosted at mcp.synapse.org on Sage Bionetworks infrastructure, no local install needed
• OAuth2 browser login means the agent inherits the user's existing Synapse permissions
• Covers the most common read operations (entity, annotations, provenance, children, search) in a small, focused tool set
• Open-source companion repo at susheel/synapse-mcp for self-hosting

• Read-only: no tools for uploading files, writing annotations, or modifying entities
• Source repo is maintained by an individual contributor rather than the Sage Bionetworks org, even though the service is hosted on the Synapse domain
• Sending Synapse data to external LLM providers may conflict with Synapse Terms of Service for controlled-access datasets; self-hosting is recommended for sensitive data

• nf-osi/mcp-server: NF-OSI's MCP server with tools tailored to Synapse data curation workflows for the Neurofibromatosis community
• Direct use of the Synapse Python client (synapseclient) or REST API for scripted access outside an MCP context
• CData Azure Synapse MCP server: unrelated (Azure Synapse Analytics warehouse, not Sage Bionetworks Synapse), listed only to disambiguate