- CollectionMeeting & Conversation
- Classificationagent-native
- LicenseMIT
- MCPAvailable
- Agent SkillUnavailable
- InterfacesMCP
- Verified2026-10-20
What it does
joinly.ai puts an AI participant into Zoom, Google Meet, or Microsoft Teams (browser-based). Official README: live voice/chat interaction, interruption-aware conversational flow, bring-your-own LLM, pluggable STT/TTS (Whisper/Kokoro local defaults; Deepgram/ElevenLabs optional). Self-host via Docker (~2.3 GB with browser + models); CUDA image available.
Distinct from catalog peers:
| Peer | Difference |
|---|---|
| Vexa | Self-host transcription + interactive bot API (17 MCP tools, REST). joinly is middleware: any agent talks MCP to a meeting connector |
| AgentCall | Hosted join-meeting Skill (ak_ac_ keys), not an OSS MCP server you run |
Agent-facing interfaces
- Interaction pattern: CLI + MCP server
- Catalog protocols: MCP
- MCP: ✅ Available — MCP is the backbone (homepage: “MCP: The backbone of joinly”)
- Agent Skill: ⚠️ No official npx skills add package published yet.
- Onboarding type:
mcp
How agents onboard
Interaction pattern: CLI + MCP server
docker pull ghcr.io/joinly-ai/joinly:latest
# Quickstart client-in-container (no external MCP):
docker run --env-file .env ghcr.io/joinly-ai/joinly:latest --client <MeetingURL>
# Server mode — bind localhost only:
docker run -p 127.0.0.1:8000:8000 ghcr.io/joinly-ai/joinly:latest
uvx joinly-client --env-file .env <MeetingUrl>
See the full onboarding section in the dossier.
Protocol surface
| Interface | Detail |
|---|---|
| Docker MCP | ghcr.io/joinly-ai/joinly:latest (--client or server :8000) |
| Client | uvx joinly-client / PyPI joinly-client |
| Extra tools | --mcp-config JSON (mcpServers) |
| Cloud | Former optional host cloud.joinly.ai — DNS dead as of 2026-09-08; not required |
Verification
Last verified 2026-10-20.
Sources:
- https://api.github.com/repos/joinly-ai/joinly