agora-moss

Moss semantic search for Agora Conversational AI, exposed as a single MCP tool (search_knowledge_base) over streamable HTTP.

Drop it behind a public URL, add one mcp_servers entry to your ConvoAI join body, and your voice agent can search your Moss index. Nothing else changes — the package never owns the LLM hop.

This is the Agora analog of vapi-moss and elevenlabs-moss.

Install

pip install agora-moss
# or
uv add agora-moss

Quickstart

import uvicorn
from agora_moss import MossAgoraSearch, create_mcp_app

search = MossAgoraSearch(
    project_id="moss-project-id",
    project_key="moss-project-key",
    index_name="my-index",
    top_k=5,
    alpha=0.8,
)

mcp = create_mcp_app(search)
app = mcp.streamable_http_app()

uvicorn.run(app, host="0.0.0.0", port=8080)

Expose the server at a public URL, then point Agora ConvoAI at it by adding this to your join body:

"llm": {
  "mcp_servers": [{
    "name": "moss",
    "endpoint": "https://<public-host>/mcp",
    "transport": "streamable_http",
    "allowed_tools": ["search_knowledge_base"]
  }]
},
"advanced_features": { "enable_tools": true }

Agora constraints (verified April 2026):

• Server entry name: ≤48 chars, alphanumeric only (no hyphens / underscores).
• Transport: streamable_http only.
• advanced_features.enable_tools must be true.

Public API

Demo app

See apps/agora-moss for a runnable end-to-end demo (Docker image + start_agent.py that calls ConvoAI's join REST).

Dependencies

• moss>=1.0.0
• mcp>=1.2 (Model Context Protocol Python SDK, FastMCP)
• Python >=3.10,<3.15

License

BSD-2-Clause. See repository root.