Developers · For AI agents

Built for AI agents. Friendly to humans too.

Drop an MCP server into Claude Desktop or Cursor. Paste an OpenAPI URL into your custom GPT. Or just curl. Your agent gets every employee + 905 media models in seconds, on a single credit balance.

Get an API key →Read the docs →Download OpenAPI →

Pick your path.

Path 1 — fastest

MCP server

Claude Desktop, Cursor, Continue, and any other MCP-aware client. Drop the snippet below into your mcp.json.

Get config →Path 2 — agentic

Custom GPT / Action

Paste our OpenAPI URL into ChatGPT's GPT builder or any tool-calling agent. The schema teaches it every endpoint.

Get spec →Path 3 — classic

Direct HTTPS

Plain JSON over HTTPS. Any HTTP client. Bring your own SDK or call us with curl.

See examples →

Path 1 — MCP server

Claude Desktop, Cursor, anything MCP-aware

Open your client's mcp.json config and paste this. Restart the client. Your AI assistant now has tools to call every employee, run any media model, check balance, and trigger autonomous agents.

{
  "mcpServers": {
    "abn": {
      "url": "https://automatebusiness.com/api/v1/mcp",
      "headers": {
        "Authorization": "Bearer abn_YOUR_KEY_HERE"
      }
    }
  }
}

On first call the assistant uses your bearer key. The server is content-routed (HTTP+SSE), no local install required.

Path 2 — OpenAPI

Custom GPT, Claude Action, agent frameworks

Our full API surface is published as an OpenAPI 3.1 document. Point your agent builder at it; it auto-generates tools.

https://automatebusiness.com/api/v1/openapi

ChatGPT GPT builder: Configure → Actions → Import URL → paste the link above. Set the auth as API Key, Bearer, paste your abn_… token.

Path 3 — direct HTTPS

Just curl (or any HTTP client)

No SDK lock-in. Plain JSON, plain HTTPS. Replace abn_YOUR_KEY_HERE with a key from /app/api-keys.

Run a text employee

curl https://automatebusiness.com/api/v1/chat \
  -H "Authorization: Bearer abn_YOUR_KEY_HERE" \
  -H "Content-Type: application/json" \
  -d '{
    "employeeSlug": "copywriter",
    "prompt": "Write 3 headlines for a plumbing landing page"
  }'

Generate an image

curl https://automatebusiness.com/api/v1/models/run \
  -H "Authorization: Bearer abn_YOUR_KEY_HERE" \
  -H "Content-Type: application/json" \
  -d '{
    "modelId": "openai/gpt-image-2/text-to-image",
    "params": { "prompt": "A pixel-art mushroom on cream paper" }
  }'

Synthesize a voice (text-to-speech, streaming)

curl https://automatebusiness.com/api/v1/voice/synthesize \
  -H "Authorization: Bearer abn_YOUR_KEY_HERE" \
  -H "Content-Type: application/json" \
  -o spoken.mp3 \
  -d '{
    "text": "Hi there, thanks for calling Hearth Coffee.",
    "voice": "warm-female",
    "format": "mp3",
    "model": "fast"
  }'

Catalog voices: warm-female, deep-male, energetic-female, british-male, young-female, podcast-host. Or pass a 20-char voice id directly. Stream returns audio/mpeg or audio/L16 (PCM 16kHz) for voice-agent pipelines. Billed at 1 credit per 50 characters (model: "fast") or 1 credit per 25 characters (model: "quality").

Transcribe audio (speech-to-text)

curl https://automatebusiness.com/api/v1/voice/transcribe \
  -H "Authorization: Bearer abn_YOUR_KEY_HERE" \
  -F "audio=@recording.mp3" \
  -F "language=en"

Returns text + word-level timestamps. Accepts mp3, wav, m4a, ogg, webm up to 25 MB. Add -F "diarize=true" for speaker labels. Billed at 1 credit per 2 seconds of audio.

Full endpoint reference at /docs.

Bonus

Drop-in system prompt for your agent

Copy-paste this into your agent's system message so it knows exactly how to call us. Works in Claude, ChatGPT, Cursor, or any agent runtime that accepts a system prompt.

You can call the automatebusinessnow API at https://automatebusiness.com/api/v1.
Auth: pass `Authorization: Bearer abn_YOUR_KEY_HERE` on every request.
Available endpoints:
  POST /chat                — text employees (copywriter, consultant, etc.)
  POST /images              — generate an image
  POST /videos              — generate a short video clip
  POST /voice/synthesize    — text-to-speech (streaming, voice-agent ready)
  POST /voice/transcribe    — speech-to-text (mp3/wav/m4a, word timestamps)
  POST /audio               — text-to-music
  POST /upscale             — upscale an image
  POST /models/run          — run any of 905+ catalog models by id
  GET  /credits             — current credit balance
  GET  /status              — public health check (no auth)

Every response is JSON shaped { "ok": true/false, "creditsUsed": ..., "balance": ... }.
Failed jobs auto-refund. Per-key rate limits return 429.
Full reference: https://automatebusiness.com/docs

How billing works

One credit balance for everything — web app, autonomous agents, API. View at /app/credits.
Failed jobs auto-refund. Every successful response includes creditsUsed and the new balance.
Top-ups never expire. Plan credits refresh monthly. Free signup tier: 200 credits, no card.
Per-key safety.A single key burning > 100,000 credits in 60 seconds ($1,000/min sustained) is auto-revoked. That ceiling sits comfortably above the most expensive single call (a Veo 3 video is ~1,800 credits) — it's a runaway-loop circuit, not a per-call cap. Read more.

Questions? Reply to hello@automatebusiness.com. Real human, one business day max.