synapse mcp server
Basic MCP server that connects AI clients to your Synapse backend API.
tools exposed
- health_check
- search_memories
- list_recent_memories
- get_memory
- create_memory
- ask_synapse
setup
- Install dependencies
cd /home/pranav/work/synapse/mcp
pip install -r requirements.txt- Configure environment
cp .env.example .envSet values in .env:
- SYNAPSE_API_URL: backend URL, example http://localhost:8000
- SYNAPSE_API_TOKEN: user JWT token from Synapse login
- SYNAPSE_API_TIMEOUT: optional, default 15 seconds
- Run server
python server.pybasic manual verification
After attaching this MCP server in your client (Claude/Cursor/VS Code), call tools in this order:
- health_check
- Expected: backend ok and token_configured true.
- list_recent_memories with limit 3
- Expected: ok true and memory list in data.
- search_memories with a known query
- Expected: ok true and search results in data.
- create_memory
- Example inputs:
- memory_type: text
- title: MCP smoke memory
- text_content: created from mcp
- Expected: ok true and created memory in data.
- get_memory with the id returned from create_memory
- Expected: same memory details.
- ask_synapse with a simple question
- Expected: ok true and reply payload in data.
You can also call ask_synapse with use_internet: true to let the backend add web context to the answer when needed.
If any tool returns ok false:
- Check SYNAPSE_API_URL
- Check SYNAPSE_API_TOKEN validity
- Check backend logs for endpoint errors
claude desktop config example
{
"mcpServers": {
"synapse": {
"command": "python",
"args": ["/home/pranav/work/synapse/mcp/server.py"]
}
}
}
