What it does
This MCP server gives a client access to Fish Audio text-to-speech models via the AceDataCloud platform. It generates speech from text, browses, searches and fetches metadata for Fish voice models, submits generation tasks that are polled for results, and can look up several task results in one call. A one-shot option conditions a single request on a public HTTPS reference audio URL and its exact transcript without creating a reusable voice. Saved or public voices can instead be referenced by their ID.
It suits workflows that need spoken audio generated from text, or that should reuse a Fish voice from the public library or from voices the calling account has saved. Listing saved voices is filtered to the calling account, supports sorting by creation date or task count, and paginates results. Saved voices are public and can be searched and used by others. Billing follows the UTF-8 byte count of the target text.
The server can be reached through hosted OAuth at the remote endpoint, through the same endpoint with an AceDataCloud API token sent as a Bearer header, or through a local stdio process that reads an API token from an environment variable and still calls the AceDataCloud API. OAuth requires an AceDataCloud account and consent in the browser. API tokens should be kept in a secret store or environment variable. OAuth and a fixed Authorization header should not both be configured for the same server.
Limitations and requirements
Requires an AceDataCloud account and an API key or OAuth sign-in. The hosted service can be metered, so current pricing should be checked before running real operations. Reference audio must be a public HTTPS URL, in MP3 or WAV format, 10 to 270 seconds long, with its exact transcript. One-shot reference fields and reference_id cannot be combined. Saved voices are public and can be used by others. Dynamic Client Registration registers the client application and is not an API key. Task lookup does not grant permission to submit generation. A platform account-management token cannot replace a generation API key. Local stdio requires the ACEDATACLOUD_API_TOKEN environment variable.
Safety
Our team has not reviewed this resource. Only automated rules checked it; resources with high-risk signals are not published. These are signals, not a guarantee that installing it is safe. Try it in a sandbox project first.
- Runs code or changes an assistant's behaviour
Sources
- Official MCP Registry first seen 13 Nov 2026 · last seen 15 Nov 2026 · its page on the source
io.github.AceDataCloud/mcp-fish
MCP server exposing the AceDataCloud Fish Audio API (text-to-speech with voice conditioning)
v2026.11.15.0As shown on registry.modelcontextprotocol.io when we last read it