joinly-client 0.1.6

Client for joinly: Make your meetings accessible to AI Agents

joinly-client: Client for a conversational meeting agent used with joinly

Prerequisites

Set LLM API key

Export a valid API key for the LLM provider you want to use, e.g. OpenAI:

export OPENAI_API_KEY="sk-..."

Or, create a .env file in the current directory with the following content:

OPENAI_API_KEY="sk-..."

For other providers, export the corresponding environment variable(s) and set provider and model with the command:

uvx joinly-client --llm-provider <provider> --llm-model <model> <MeetingUrl>

Start joinly server

Make sure you have a running joinly server. You can start it with:

docker run -p 8000:8000 ghcr.io/joinly-ai/joinly:latest

For more details on joinly, see the GitHub repository: joinly-ai/joinly.

Command line usage

We recommend using uv for running the client, you can install it using the command in their repository.

Connect to a running joinly server and join a meeting, here loading environment variables from a .env file:

uvx joinly-client --joinly-url http://localhost:8000/mcp/ --env-file .env <MeetingUrl>

Add other MCP servers using a configuration file:

{
    "mcpServers": {
        "localServer": {
            "command": "npx",
            "args": ["-y", "package@0.1.0"]
        },
        "remoteServer": {
            "url": "http://mcp.example.com",
            "auth": "oauth"
        }
    }
}

uvx joinly-client --mcp-config config.json <MeetingUrl>

You can also set other session-specific settings for the joinly server, e.g.:

uvx joinly-client --tts elevenlabs --tts-arg voice_id=EXAVITQu4vr4xnSDxMa6 --lang de <MeetingUrl>

For a full list of command line options, run:

uvx joinly-client --help

Code usage

Direct use of run function:

import asyncio

from dotenv import load_dotenv
from joinly_client import run

load_dotenv()


async def async_run():
    await run(
        joinly_url="http://localhost:8000/mcp/",
        meeting_url="<MeetingUrl>",
        llm_provider="openai",
        llm_model="gpt-4o-mini",
        prompt="You are joinly, a...",
        name="joinly",
        name_trigger=False,
        mcp_config=None,  # MCP servers configuration (dict)
        settings=None,  # settings propagated to joinly server (dict)
    )


if __name__ == "__main__":
    asyncio.run(async_run())

Or only using the client and a custom agent:

import asyncio

from joinly_client import JoinlyClient
from joinly_client.types import TranscriptSegment


async def run():
    client = JoinlyClient(
        url="http://localhost:8000/mcp/",
        name="joinly",
        name_trigger=False,
        settings=None,
    )

    async def on_utterance(segments: list[TranscriptSegment]) -> None:
        for segment in segments:
            print(f"Received utterance: {segment.text}")
            if "marco" in segment.text.lower():
                await client.client.call_tool("speak_text", {"text": "Polo!"})

    client.add_utterance_callback(on_utterance)

    async with client:
        # optionally, load all tools from the server
        tool_list = await client.client.list_tools()

        await client.join_meeting("<MeetingUrl>")
        try:
            await asyncio.Event().wait()  # wait until cancelled
        finally:
            print(await client.get_transcript())  # print the final transcript


if __name__ == "__main__":
    asyncio.run(run())