Crikk MCP

Give your AI assistant a voice.

Turn a script into speech from your conversation. Connect Crikk to ChatGPT, Claude, or another MCP client to browse saved recordings, find voices, and create MP3 audio in your existing library.

Claude opens a prefilled connector form for you to review. ChatGPT requires manual setup. See connection instructions.

Connect using your Crikk account. No API key needed.

From a prompt to an MP3

Example request

Use Crikk to find an English voice, turn this sentence into speech, save it in my Home library, and give me the audio link: “Welcome to our next chapter.”
  1. 1. Choose a voice
  2. 2. Generate speech
  3. 3. Save the MP3

Start here

Connect your Crikk account

MCP lets an AI assistant use Crikk’s speech tools on your behalf. Add the server URL below in your client, sign in to your Crikk account, and approve the connection. You’ll need a client that supports custom remote MCP servers.

MCP server URL

https://crikk.com/mcp

Transport: Streamable HTTP. Paste this URL into your client’s server URL field.

ChatGPT

Connect with OAuth

Add Crikk manually using this server URL:

https://crikk.com/mcp

  1. Open ChatGPT Plugins(opens in a new tab), select the plus button, then Add custom MCP server. Availability depends on your plan and workspace permissions.
  2. Add a connection named Crikk with the server URL above.
  3. Select OAuth and automatic / dynamic client registration. Crikk does not currently support the published identity (CIMD) option.
  4. Sign in to your Crikk account when prompted. On the authorization page, review speech generation and library access, then choose Connect Crikk.
  5. Create and install the plugin, then enable Crikk in your conversation and try the prompt below.
OpenAI’s connection authentication guide

Claude

Connect with OAuth

The Add to Claude button prefills Crikk’s name and server URL. If the form opens empty after sign-in or on mobile, enter them using the steps below.

  1. Open Customize → Connectors(opens in a new tab) and add a custom connector. On a team workspace, an administrator may need to add it first.
  2. Name it Crikk and enter the server URL above.
  3. Choose Sign in now and Register automatically for the OAuth client.
  4. Sign in to your Crikk account when prompted, then approve speech generation and library access on Crikk’s authorization page.
  5. Enable Crikk for your conversation and ask it to generate speech.
Claude’s custom connector guide

Sign in only on Crikk’s website. Your assistant receives separate MCP access tokens; it does not receive your Crikk login credentials.

Try your first voiceover

After connecting, paste this prompt into a conversation with Crikk enabled.

Use Crikk to find an English voice, turn this sentence into speech, save it in my Home library, and give me the audio link: “Welcome to our next chapter.”

You can also ask: “List my projects,” “Show the audio in Home,” or “Save this voiceover in one of my projects.” Choose a different language, speed, or a supported speaking style when generating.

Your audio library, in one conversation

Your assistant can browse existing recordings or create new speech in Home or a selected project.

1

Browse your projects

list_projects

List your existing projects, including Home. Use the returned project ID to choose a folder for listening or generation. Does not use speech characters.

Inputs: none

2

Listen to saved audio

list_audio

Find existing recordings with playback and download URLs, newest first. Returns up to 20 files per page; request the next page when has_next_page is true. Does not use speech characters.

Inputs: optional project_id (default home), page (default 1)

3

Find your voice

list_voices

Browse the free voice catalog, languages, and supported speaking styles. Filter by a language or locale such as English or en-US.

Inputs: language (optional)

4

Create and save audio

generate_speech

Generate speech and automatically save it to your existing Crikk library. Omit project_id or use home for Home, or pass an ID from list_projects. Returns a job ID.

Inputs: text, voice_name; optional project_id, speaking_style, speed, pitch

5

Get the saved MP3

get_speech_result

Check the same job until it completes. Returns the saved recording’s ID, project ID, title, text, voice, audio_url, and MIME type. Repeated checks do not charge again.

Inputs: job_id

Usage and downloads

API and MCP access includes only Crikk’s free voice catalog. Premium voices are not available through these integrations. Speech generation uses your account’s API character allowance.

12,000 spoken characters per request
Split longer scripts into separate generations. MCP uses your account’s existing API character allowance.
Saved to your existing library
Home is the default destination. Pass a project ID from list_projects for another folder. Use the returned audio_url to listen or download later, under normal library retention rules.
Shared generation rules
MP3 output is fixed at 48 kHz, 192 kbps, mono for API and MCP voices. Pauses such as [2s] and [500ms] are capped at 10 seconds each. The shared Microsoft request timeout is 50 minutes. Output format, timeout, and maximum pause are server settings; do not send them as tool inputs.
Polling and playback
Job records last ten minutes from submission. Completed audio remains in the library after the polling record expires. get_speech_result returns JSON with an audio_url for playback and download.
Adjust the delivery
Speed: 0.5× to 2×. Pitch: −50% to +50%. Speaking styles depend on the selected voice. Default speed is 1× and pitch is unchanged.
One generation, one charge
A completed recording uses characters once when saved, even without polling. Checking a job or listing existing audio does not charge again.

Questions and troubleshooting

How do I sign in?

Sign in to your Crikk account using your usual login method, then approve the connection. Your assistant receives separate MCP access tokens. You do not need an API key or numeric user ID.

Does MCP have a separate allowance?

No. MCP shares your API character allowance. Characters are charged once when a completed recording is saved, even if you never poll the job. Listing projects, browsing recordings, and checking results do not use additional speech characters.

Where does generated audio go?

Into your existing Crikk library, alongside dashboard and mobile recordings. Home is the default. To use another project, ask your assistant to list projects and use the returned project ID. Audio is saved automatically when generation completes.

What if the audio is still generating?

Ask your assistant to check the same job again. Polling records last ten minutes from submission and can be lost on a server restart. Their expiry does not delete saved audio. Use list_audio to find a completed recording later before generating the same text again.

Why can’t I see my saved files?

Connections need library:read permission to list projects and recordings. Reconnect Crikk and approve library access if your existing connection only has speech:generate.

Why can’t I connect?

Check the server URL and sign in to Crikk when prompted. Use OAuth with automatic client registration rather than published client identity (CIMD). Your client plan and workspace permissions must allow custom connections.

Do downloads still expire after one hour?

The old temporary MCP download flow has been removed. Results now link to saved library recordings under the normal library retention rules. Use list_audio to find a saved recording and its current audio URL.

How do I disconnect?

Remove the connection in your AI app. MCP connections use their own authorization and expire after up to 30 days, after which you need to reconnect.