| name | voice-join |
| description | [Pro] Join a conference call as a voice AI assistant that can answer questions about the current codebase. Dials into Zoom, Meet, or Teams via phone. |
| user-invocable | true |
| disable-model-invocation | true |
| allowed-tools | ["mcp__codebase-voice__start_call_session","mcp__codebase-voice__poll_question","mcp__codebase-voice__provide_answer","mcp__codebase-voice__stop_session","mcp__codebase-voice__get_config","Read","Grep","Glob"] |
Voice Join — Join a Call as a Codebase AI Assistant [Pro]
Join a conference call via SIP dial-in. The assistant dials the conference number, joins as its own participant ("Claude - Codebase Assistant"), and answers codebase questions from anyone on the call.
Requires LiveKit + Twilio configuration. Run /voice-config first.
Arguments
$1 — The dial-in phone number (e.g., +15551234567)$2 — The meeting PIN/access code (optional, e.g., 12345#)
If no arguments are provided, ask the user for the dial-in number.
Steps
-
Check configuration: Call mcp__codebase-voice__get_config to verify Deepgram AND telephony keys (LiveKit + Twilio) are set. If telephony is not configured, tell the user this is a Pro feature, suggest /voice-start as the free alternative, and stop.
-
Index the codebase: Quickly scan the current project:
- Use
Glob with **/*.{py,js,ts,go,rs,java,rb,cpp,c,h} to get source files (limit to first 50)
- Use
Read to read: README.md, CLAUDE.md, package.json or pyproject.toml
- Build a 2-3 sentence codebase summary
-
Join the call: Call mcp__codebase-voice__start_call_session with the project path, codebase summary, dial-in number, and optional PIN.
-
Tell the user the assistant has joined the call.
-
Enter the question-answer loop: Same as /voice-start:
LOOP:
a. Call mcp__codebase-voice__poll_question (timeout_ms: 2000)
b. If result is "__NO_QUESTION__" → go to step a
c. If result is "__LEAVE__" → call stop_session and exit
d. If result is "__SESSION_ENDED__" → exit
e. Otherwise, it's a codebase question:
- Use Grep and Read to find relevant code
- Formulate a concise answer (1-3 sentences, conversational)
- Call mcp__codebase-voice__provide_answer with the answer
- Go to step a
-
On exit: Call mcp__codebase-voice__stop_session and tell the user the session ended.
