Does LIvekit has a MCP I can connect Claude to inspect the sessions, livekit logs...

Thanks for reaching out, Yousri. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.

• I'm trying to connect Claude to LIvekitCloud • I want Claude to analyze stored session logs. • I don't care about documnetation. I want Claude to be able to fetch a given session ID and inspect the logs

Thanks — this is helpful. You want Claude (an external assistant) to fetch stored session logs from LiveKit Cloud by session ID and analyze them. Here’s a clear, practical way to do that and the questions I need to give exact steps. What you can do (high level) • LiveKit Cloud exposes stored agent/session logs through Agent Observability; those logs can be accessed from the Cloud Dashboard, the LiveKit CLI, or forwarded to external logging services (Datadog, Sentry). • To let Claude fetch logs programmatically, the assistant needs an API or endpoint that returns the session logs for a given session ID. You can provide that by: ◦ Using the Cloud Dashboard/CLI to export logs for a session, then making those exported logs available to Claude via an API endpoint or secure file storage the assistant can call, OR ◦ Forwarding LiveKit logs to an external logging service (Datadog/Sentry) that Claude can query (if Claude can be given API access to that external service). Recommended approach (practical, secure) 1. Decide where Claude will query logs from: ◦ Option A — LiveKit Cloud -> Export/CLI endpoint: ▪︎ Use the Cloud Dashboard or LiveKit CLI to fetch stored logs/transcripts for a specific session ID and expose them via a small internal API (authenticated) that Claude can call. ◦ Option B — Forward logs to an external logging datastore: ▪︎ Configure LiveKit to forward logs to Datadog or Sentry. Give Claude scoped API credentials for that logging service and a query that returns logs matching the session ID. 1. Implement a minimal, authenticated API (if you choose Option A): a. Create an internal endpoint (e.g., GET /sessions/:sessionId/logs) that: ▪︎ Authenticates the caller with a short-lived token or scoped API key. ▪︎ Looks up the session logs from Cloud exports or from whatever storage you use for archived logs. ▪︎ Returns structured logs, transcripts, and any related traces in JSON. a. Give Claude the endpoint URL and a scoped API key with only read permission for logs. b. Limit returned data to the session ID requested to reduce exposure. 1. If you choose the external-logs route (Option B): a. Forward LiveKit logs to the external service (Datadog or Sentry). b. Create a minimal query or saved search that fetches logs for a given session ID. c. Give Claude an API key with read-only access and the exact query/endpoint to call. Security and operational notes • Use scoped, short-lived credentials for Claude rather than broad admin keys. • Restrict the assistant’s access to only the specific logs it needs (by session ID). • Return structured JSON (logs, timestamps, transcripts, trace IDs) so Claude can parse and analyze without extra conversion. Follow-up questions so I can give exact commands and code • Do you already forward LiveKit logs to Datadog/Sentry, or do you plan to use the Cloud Dashboard/CLI export route? • Can you host a small authenticated API endpoint for Claude to call, or do you need an example of how to fetch logs via the CLI and expose them programmatically? • Do you want example code (Node/Python) for an endpoint that returns session logs in JSON and uses a scoped API key? Tell me which option you prefer (Cloud export + API or external logging service) and whether you want code examples — I’ll provide step-by-step commands and a small sample implementation for that path.