API Access

One API for 500+ AI models The Krater API is a REST API at https://api.krater.ai/v1 that gives you programmatic access to everything in Krater: chat with 500+ AI models, generate images and video, ...

One API for 500+ AI models

The Krater API is a REST API at https://api.krater.ai/v1 that gives you programmatic access to everything in Krater: chat with 500+ AI models, generate images and video, text-to-speech, transcription, music, web search, AI detection, plagiarism checks, code, and documents — all billed in Krater credits from your plan.

Requests and responses follow OpenAI-compatible shapes wherever an OpenAI equivalent exists, so existing OpenAI SDKs work by changing the base URL. Full endpoint reference, examples, and error documentation live in the API docs.

Who gets API access

API access is included on Ultra and Max plans.

• Past-due or suspended accounts lose access until billing is resolved.
• Upgrade any time in Settings → Plan. See plans overview.

Create an API key

1. Open Settings → API
2. Click Create Key and give it a name (e.g. “My Python Script”)
3. Copy the key immediately — it is shown only once and stored as a hash

You can have up to 5 active keys. Keys look like kr_live_... and are sent in the Authorization header:

Authorization: Bearer kr_live_your_key

Revoke a key from the same panel at any time — revocation is immediate and permanent. Keys created by third-party apps over OAuth show a Connected app badge; revoking one also disconnects that app’s access.

Your first call

curl https://api.krater.ai/v1/chat/completions \
  -H "Authorization: Bearer kr_live_your_key" \
  -H "Content-Type: application/json" \
  -d '{"model":"openai/gpt-4o-mini","messages":[{"role":"user","content":"Hello!"}]}'

With the OpenAI Python SDK, just repoint the client:

from openai import OpenAI

client = OpenAI(
    api_key="kr_live_your_key",
    base_url="https://api.krater.ai/v1"
)

response = client.chat.completions.create(
    model="openai/gpt-4o-mini",
    messages=[{"role": "user", "content": "Hello!"}]
)

Chat completions support streaming: pass "stream": true and responses arrive token-by-token over Server-Sent Events.

Endpoints

GET /v1/models costs no credits — use it to discover model IDs programmatically.

Credits and billing

Every API call spends credits from the same pool as your in-app usage — there is no separate API balance.

• Chat, code, and document endpoints bill on actual token usage at each model’s rate.
• Media and utility endpoints have per-call costs listed in the API docs.
• Track spend in Settings → Usage, and top up on Ultra or Max if you need more. See understanding credits.

Limits

• Exceed the per-minute limit and you get 429 with a rate_limit_error — back off and retry.
• Exceed the daily cap and you get 402 until the window resets.
• Need a higher per-key cap or throughput? Contact hey@krater.ai.

Errors

All errors return a consistent JSON shape with a machine-readable type, so 401, 402, 429, and upstream failures are easy to handle in code. The API docs cover the full format, and common error codes explains what each code means.

Prefer MCP over REST?

If you want Claude, ChatGPT, Cursor, or another MCP client to call Krater tools directly — without writing against REST endpoints — use the Krater MCP server instead:

https://api.krater.ai/v1/mcp

It authenticates with OAuth sign-in and exposes 20 tools (chat_completion, generate_image, generate_video, text_to_speech, web_search, list_models, and more), billed against your plan credits. Connections appear in Settings → API as connected apps and can be revoked there.

Related

• Understanding Credits
• Plans Overview
• Common Error Codes
• Developers page — SDK examples and the full API surface

More in Features

• Additions: Apps, Personas & Prompts
• Personas: Specialist AI Assistants
• Scheduled Tasks
• Sharing & Projects
• Krater MCP Server

See Krater pricing or contact support.