Context
The MCP cognition server (tools/mcp_server.py) needs real-world testing with actual MCP clients to validate the tools work correctly in practice.
Test matrix
| Client |
Transport |
Priority |
| Claude Desktop |
streamable-http |
High |
| Cursor |
streamable-http |
High |
| Letta Code (converting-mcps-to-skills) |
stdio |
Medium |
| Custom Python client |
stdio |
Low |
Test scenarios
- Search:
search_cognition("coordination patterns") returns results from multiple agents
- Read:
read_agent_cognition("void.comind.network", "stream.thought.memory") returns void's recent thoughts
- Write:
write_thought("test thought") creates a record on ATProto
- Discovery:
list_indexed_agents() returns stats and DID list
- Error handling: Missing credentials, indexer down, invalid handles
Depends on
Context
The MCP cognition server (
tools/mcp_server.py) needs real-world testing with actual MCP clients to validate the tools work correctly in practice.Test matrix
Test scenarios
search_cognition("coordination patterns")returns results from multiple agentsread_agent_cognition("void.comind.network", "stream.thought.memory")returns void's recent thoughtswrite_thought("test thought")creates a record on ATProtolist_indexed_agents()returns stats and DID listDepends on