Agent access (MCP)

Open Translator can mirror the text of your transcripts and the speaker names you label to the cloud and serve them, read-only, to any AI agent that speaks the Model Context Protocol. Audio and images never leave your device. Sync is off until you turn it on.

1. Turn on sync and create a token

1. Sign in (guest accounts cannot sync).
2. Settings → Account → Agent access (MCP).
3. Turn on Sync transcripts for agents. Existing sessions upload once.
4. Tap New token, name it after the agent, copy the token. It is shown only once; revoke it any time from the same screen.

2. Connect your agent

After creating a token the app shows a setup prompt. Copy it, or share it to the device your agent runs on, and paste it into the agent's chat: Claude Code, Codex, Pi, Hermes, OpenClaw or any other MCP-capable agent. The prompt carries the server URL, the token and a description of the tools; the agent adds the server to its own configuration and confirms by listing your sessions.

Setting it up by hand instead: transport Streamable HTTP, URL as shown in the app, header Authorization: Bearer ot_….

3. What the agent can do

• list_transcripts — your sessions, newest first; is_live marks one still being recorded.
• get_transcript — one session as bilingual text or JSON; poll a live session with since_seq.
• search_transcripts — find what was said across recent sessions.
• Resource transcript://<session_id> for clients that attach resources.

Agents can only read. Turning sync off deletes the cloud copy; deleting your account deletes everything, tokens included.