mcp.film Submit

mcp.film / Voice & Dialogue / Fish Audio MCP

Fish Audio MCP

Hosted Fish Audio server: expressive TTS, cloning, transcription, plus quoted image, video and lip-sync jobs.

OfficialHosted remoteOAuthfreemiumverified Site ↗Docs ↗

MCP server URL
https://api.fish.audio/mcp

Caveats

  • Hosted only and closed source, and the tool names come from the announcement and docs, not a live tools/list.
  • Every image or video job is quote, approve, then submit, and the server reprices at submission; if the price moved, quote again.
  • Video, lip sync and avatars need Plus or higher, and the free tier is personal, non-commercial use only.
  • Generated outputs come back as permanent URLs, so assume anything you generate is reachable by link.

What it does

Fish Audio's official hosted MCP server connects an agent to your Fish Audio account over OAuth: search the voice library and clone voices, generate speech steered with inline tags like [whispering] or [excited], transcribe local or uploaded audio, and — since September and October 2026 — generate, edit and upscale images and run text-to-video, image-to-video, lip sync and talking avatars behind a free quote step. For a filmmaker-agent the draw is temp dialogue and scratch VO with emotional direction written into the script line, then a lip-synced or avatar pass on the same account, all billed against the plan you already have.

Auth

OAuth — Browser OAuth with your Fish Audio account; you pick the team the connection works in

Connect

Connect Fish Audio MCP to Claude Code

Claude Code
claude mcp add --transport http fish-audio https://api.fish.audio/mcp

Connect Fish Audio MCP to Claude Desktop

Settings → Connectors → Add custom connector → paste https://api.fish.audio/mcp. Claude Desktop connects to hosted servers natively; the config file below is the fallback for accounts without the Connectors panel.

Fallback: claude_desktop_config.json via mcp-remote
{
  "mcpServers": {
    "fish-audio": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://api.fish.audio/mcp"
      ]
    }
  }
}

Connect Fish Audio MCP to Cursor

Cursor (.cursor/mcp.json)
{
  "mcpServers": {
    "fish-audio": {
      "url": "https://api.fish.audio/mcp"
    }
  }
}

Hosted endpoint for other MCP clients

Remote endpoint (Streamable HTTP)
https://api.fish.audio/mcp

Sample tools

search_voicescreate_voice_clonetext_to_speechspeech_to_textlist_media_modelsestimate_image_generationestimate_video_generationgenerate_imagegenerate_image_editgenerate_videoget_generation_statusget_generation_resultstudio_generatestudio_export

Field notes

Full curator notes (verified Oct 26, 2026)

Hosted only and closed source — no npm or PyPI package, no public server repo. Re-checked 2026-10-26: endpoint, Claude Code line and OAuth unchanged, but the surface grew. Correction to the earlier note here: Fish Audio's 7 Oct 2026 announcement adds image upscaling, lip sync and talking avatars, and the docs now list voice cloning, Story Studio (studio_generate, studio_export), an asset library and account tools. Lip sync and avatars run through generate_video rather than tools of their own. The tool names above come from that post and the docs page, not a live tools/list. Every image or video job goes discover, free estimate, your approval, then submit; the server reprices at submission and rejects the job if the price moved, so the agent has to quote again. Failed generations are refunded. Billing draws from your plan's package credits, not developer API credits; speech is 1 credit per UTF-8 byte and is charged on run with no quote. Video, lip sync and avatars need Plus ($11/mo promo, $15 list) or higher, Story Studio export needs a paid plan, and Fish publishes no fixed media price list — the estimate is the price; the free tier gets a daily image allowance and is personal, non-commercial use only, so anything headed for a paying client needs a paid plan. Connections are scoped to one team. Local files sent for transcription are uploaded to a temporary slot deleted after 7 days; generated outputs come back as permanent URLs, so assume anything you generate is reachable by link. Do not confuse it with Fish Audio's separate read-only docs MCP at docs.fish.audio/mcp, or with the community wrappers (da-okazaki/mcp-fish-audio-server, AceDataCloud/FishMCP), which are not maintained by Fish Audio.

Also in Voice & Dialogue

Common questions

What is the Fish Audio MCP server URL?
https://api.fish.audio/mcp, a hosted MCP endpoint over streamable HTTP. Auth: OAuth.
Does Fish Audio MCP need an API key?
No key to paste: it uses OAuth. Your MCP client opens a sign-in page the first time it connects.
How do I add Fish Audio MCP to Claude Code?
Run claude mcp add --transport http fish-audio https://api.fish.audio/mcp, then sign in when Claude Code asks (or run /mcp, pick the server and choose Authenticate).
Why does Fish Audio MCP return 401 or ask for authorization?
That is expected. On Nov 10, 2026 its endpoint answered mcp.film's anonymous check with a sign-in challenge and published OAuth metadata. Complete the sign-in in your client; tools are listed only after it.
Is Fish Audio MCP official?
Yes, Fish Audio maintains it. mcp.film last verified the listing on Oct 26, 2026.

Used in

Dub, translate and lip-sync for other languages

Pairs well with

Badge for your README

Maintain Fish Audio MCP? Link back to this listing. The badge shows the month it was last verified and updates when the curator re-checks it.

Listed on mcp.film: Fish Audio MCP, verified Oct 2026

Markdown
[![Listed on mcp.film](https://mcp.film/badge/fish-audio.svg)](https://mcp.film/mcps/fish-audio/)
HTML
<a href="https://mcp.film/mcps/fish-audio/"><img src="https://mcp.film/badge/fish-audio.svg" alt="Listed on mcp.film" height="20"></a>