How are you running AnythingLLM?

AnythingLLM desktop app

What happened?

TL; DR. I wasn't able to make a request to the AnythingLLM API that triggers the use of an MCP server. Has anyone succeeded?

I understand that to call an agent in the sense of custom skill, you need to use the API workspace/{slug}/chat (or stream-chat). I managed to do it:

curl -X 'POST' \
  'http://127.0.0.1:3001/api/v1/workspace/unicaen/chat' \
  -H 'accept: application/json' \
  -H 'Authorization: Bearer xxx' \
  -H 'Content-Type: application/json' \
  -d '{
  "message": "@agent what is the temperature in 48.929557/-0.469883 ? /exit",
  "mode": "chat",
  "sessionId": "identifier-to-partition-chats-by-external-id",
  "reset": false
}'

The response is correct (my agent always returns 10):

{
  "id": "79a1e5df-7568-4e09-97fb-14300959d693",
  "type": "textResponse",
  "sources": [],
  "close": true,
  "error": null,
  "textResponse": "The current temperature at 48.929557, -0.469883 is 10°C.",
  "thoughts": [
    "[debug]: @agent is attempting to call `agent1` tool",
    "Mona Gent-v1.0.0 called with lat:48.929557 long:-0.469883..."
  ]
}

When I ask for the level on sensor 1, which is served by MCP, nothing happens. I managed to do it with the app.

Did I miss something?

Are there known steps to reproduce?

No response