Connect with MCP

Before you start

MCP is part of some plans, and an admin can give it to a single account. If yours doesn't have it, the MCP page in the studio says so and links to the plans.

Claude

In Claude on the web or desktop, open Customize → Connectors, choose Add custom connector, name it Fazita and give it this address:

https://fazita.com/api/mcp

Click Connect, sign in to Fazita and allow it. No key is needed, and the connector works in Claude on your phone too. On a Team or Enterprise plan, an owner adds it for everyone under Organization settings → Connectors first.

Apps you connect this way are listed in MCP in the studio, where you can disconnect them.

Other clients

1. Make a key

Open MCP in the studio and make a key. It's shown once, so copy it then. You can keep up to 10 keys, one per tool, and revoke any at any time.

2. Add the server

In Claude Code:

claude mcp add --transport http fazita https://fazita.com/api/mcp \
  --header "Authorization: Bearer YOUR_KEY"

Any other client that speaks Streamable HTTP takes this in its MCP config:

{
  "mcpServers": {
    "fazita": {
      "type": "http",
      "url": "https://fazita.com/api/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_KEY"
      }
    }
  }
}

Clients that sign in with OAuth can skip the key: give them the address alone, and they'll send you to Fazita to allow them.

Make something

Ask your assistant for a video, an image or a voice-over. It calls list_models first, queues takes with create_take, then checks get_takes every 10–20 seconds until they're ready. To save explaining this each time, install the skill.

Tools

• list_models: The models your plan includes, what each offers, and what a take costs in credits.
• get_usage: Your plan, credits left, queue limits and characters left for voice.
• create_take: Queue one video or image from a prompt. Returns the take's id.
• get_takes: Check takes by id until each is ready or failed. A ready take has a download link that works for a day.
• list_takes: Your takes, from the studio or MCP: those still waiting, or the finished ones.
• list_voices: The voices a voice model offers, by language.
• create_speech: Speak a script with a voice model. Answers with the audio when it's done.

Limits

• One message per request: batches are refused.
• Takes wait and run under your plan's limits, counted with the studio's.
• Video and image takes spend credits; see what each model costs. Speech spends the voice model's characters instead.
• Video and image files are kept for 15 days, then deleted. Download the ones you want to keep.