Changelog

Follow public Vocci documentation and integration updates.

2026-08-31

MCP audio download is now generally available to every account connected through browser sign-in. A connected AI app can list a session's audio and request short-lived download links for that account's own recordings.
Added an Audio Export section to the MCP capability reference covering session_audio_list, audio_download_url, and audio_download_urls.
Documented the audio boundaries in limits and safety: 1 to 100 explicitly selected audio IDs per call, short-lived URLs, the account's own recordings only, and no raw bytes inlined in a response.

Added What Is Vocci, the canonical product description and glossary for people and AI assistants.
Defined Vocci in the /llms.txt and /llms-full.txt summaries so LLM tools understand the product before reading further.
The MCP server now announces what Vocci is at connect time through initialize instructions and a static About Vocci resource.
Context pack responses now lead with a one-line description of Vocci.

Added separate user and developer documentation paths.
Added user setup guides for ChatGPT, Claude, Claude Code, and Codex.
Prepared the Vocci MCP server URL: https://mcp.vocci.ai/mcp.
Documented OAuth setup for public clients.
Added developer guidance for MCP authentication, capabilities, search and fetch, limits, and safety.
Added LLM-friendly documentation files: /llms.txt, /llms-full.txt, and /raw/....