Build with our API.

Everything you can do in the web app you can do via the API. Same credits, same employees, same media models — exposed as HTTPS endpoints under https://automatebusiness.com/api/v1.

Your first call

The status endpoint is unauthenticated — paste it into your terminal to confirm the API is reachable from your machine.

curl https://automatebusiness.com/api/v1/status

You should see something like:

{
  "ok": true,
  "service": "automatebusinessnow",
  "version": "v1",
  "timestamp": "2026-04-25T00:00:00.000Z"
}

Now run a real job. Replace abn_YOUR_KEY_HERE with a key from your API keys page.

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 punchy headlines for a coffee shop"
  }'

Credits & billing

Every API call deducts from the same credit balance you see at /app/credits. Failed jobs auto-refund. Plan credits refresh monthly; top-up credits never expire.

Heads-up: when a single API key burns more than 100,000 credits in 60 seconds ($1,000/min sustained), the key is auto-revoked as a safety circuit. That ceiling sits well above any legitimate single-call cost — the most expensive video models run ~1,800–6,000 credits per call — but it stops a runaway loop dead before it bleeds your balance. Generate a new key from /app/api-keys if it ever trips. For vetted high-volume accounts we lift the cap — email hello@automatebusiness.com.

All /v1/* endpoints (except /v1/status) require a bearer token. You can have an unlimited number of keys per account, name them per integration, and revoke them individually.

Creating keys

Open /app/api-keys in the web app, give your key a name ("Zapier", "internal-script", etc.), and copy the resulting abn_… string. Keys are shown once at creation and never again — store them in your secret manager.

Sending the bearer

Authorization: Bearer abn_YOUR_KEY_HERE

Anything else (no header, malformed prefix, revoked key) returns 401 { "ok": false, "error": "Unauthorized" }.

Every endpoint that runs a model accepts a synchronous call, but for production workloads the async submit-then-poll patternis cleaner: submit returns instantly with a job id, you poll a status endpoint until the result is ready. Same contract Wavespeed exposes — drop-in compatible with any poll-loop you've already written.

POST /v1/jobs/submit

Submit a job. Returns 202 Accepted in ~1s with a job id and a poll URL. Credits are reserved up-front; you get a fast 402 if the balance is too low.

curl https://automatebusiness.com/api/v1/jobs/submit \
  -H "Authorization: Bearer abn_YOUR_KEY_HERE" \
  -H "Content-Type: application/json" \
  -d '{
    "modelId": "openai/gpt-image-2/text-to-image",
    "prompt": "A pixel-art mushroom",
    "aspect_ratio": "1:1",
    "quality": "medium",
    "resolution": "1k"
  }'

{
  "ok": true,
  "jobId": "abn_job_J-ZydDf8bghNUF8RYKS-nA",
  "modelId": "openai/gpt-image-2/text-to-image",
  "status": "submitted",
  "estimateCredits": 18,
  "pollUrl": "/api/v1/jobs/abn_job_J-ZydDf8bghNUF8RYKS-nA",
  "createdAt": "2026-04-25T02:07:30.467Z"
}

Body shape is flat — top-level params alongsidemodelId. The legacy nested form { "modelId": "...", "params": { ... } } is also accepted.

GET /v1/jobs/{id}

Single-shot poll. Non-blocking. Returns the current status; on completion finalizes the credit ledger atomically and returns the output URL. Subsequent polls of a completed job are idempotent — they re-read the cached terminal state, balance doesn't change.

curl https://automatebusiness.com/api/v1/jobs/abn_job_... \
  -H "Authorization: Bearer abn_YOUR_KEY_HERE"

While running:

{
  "ok": true,
  "jobId": "abn_job_...",
  "modelId": "openai/gpt-image-2/text-to-image",
  "status": "running",
  "estimateCredits": 18,
  "createdAt": "..."
}

On completion:

{
  "ok": true,
  "jobId": "abn_job_...",
  "modelId": "openai/gpt-image-2/text-to-image",
  "status": "completed",
  "outputUrl": "https://...png",
  "creditsUsed": 18,
  "balance": 18093,
  "createdAt": "...",
  "completedAt": "..."
}

On failure (credits auto-refunded):

{
  "ok": false,
  "jobId": "abn_job_...",
  "modelId": "...",
  "status": "failed",
  "error": "human-readable sentence",
  "creditsUsed": 0,
  "createdAt": "..."
}

Polling cadence: 4–8 seconds is plenty. Most image jobs land in 30–60s; videos and music in 1–4 minutes.

Every endpoint accepts and returns application/json. Responses are always shaped { ok, ... } on success and { ok: false, error } on failure.

{
  "employeeSlug": "copywriter",
  "prompt": "Write 3 cold-email subject lines for HVAC contractors",
  "imageUrl": null,
  "fileContext": null
}

{
  "ok": true,
  "employeeSlug": "copywriter",
  "kind": "text",
  "message": "1. \"Your inbox is leaking leads...\"\n2. ...",
  "imageUrl": null,
  "provider": "ai",
  "creditsUsed": 2,
  "balance": 18119
}

{
  "prompt": "Minimalist logo for a coffee shop named Hearth",
  "aspect_ratio": "1:1"
}

{
  "ok": true,
  "outputUrl": "https://...png",
  "creditsUsed": 8,
  "balance": 18111
}

{
  "prompt": "Pixel-art cat walking across a desk",
  "duration": 5
}

{
  "ok": true,
  "outputUrl": "https://...mp4",
  "creditsUsed": 150,
  "balance": 17961
}

{
  "text": "Welcome to Hearth Coffee. We open at 7.",
  "voice": "alloy"
}

{
  "ok": true,
  "outputUrl": "https://...mp3",
  "creditsUsed": 4,
  "balance": 17957
}

{
  "text": "Hi, thanks for calling. How can I help you today?",
  "voice": "warm-female",
  "format": "mp3",
  "model": "fast"
}

Streamed audio bytes (Content-Type: audio/mpeg or audio/L16).
Headers: X-Credits-Used, X-Voice-Format.

Catalog voices: warm-female, deep-male, energetic-female,
  british-male, young-female, podcast-host.
  Or pass a 20-char raw voice id directly.
Formats: mp3 (default) | pcm_16000 (telephony pipelines).
Models:  fast (~300ms TTFB) | quality (better prosody, slower).
Billing: 1 credit per 50 chars on fast,
         1 credit per 25 chars on quality.

# multipart/form-data
audio:    <binary mp3/wav/m4a/ogg/webm>
language: en          # optional — auto-detect if omitted
diarize:  true        # optional — speaker labels per word

{
  "ok": true,
  "text": "the full transcript here",
  "words": [
    { "word": "the",  "start": 0.12, "end": 0.18 },
    { "word": "full", "start": 0.18, "end": 0.42 }
  ],
  "languageCode": "en",
  "durationSeconds": 12.4,
  "creditsUsed": 7
}

Billing: 1 credit per 2 seconds of audio (~$18/hr retail).

# Before (direct provider)
POST https://<provider>/v1/text-to-speech/{voice_id}/stream
Headers:  xi-api-key: <provider-key>
Body:     { text, model_id: "<provider-model>", voice_settings: {...} }

# After (through automatebusinessnow)
POST https://automatebusiness.com/api/v1/voice/synthesize
Headers:  Authorization: Bearer abn_live_xxx
Body:     { text, voice: "warm-female", format: "mp3", model: "fast" }

# Before
POST https://<provider>/v1/speech-to-text
Headers:  xi-api-key: <provider-key>
Form:     file=<bytes>, model_id="...", timestamps_granularity=word

# After
POST https://automatebusiness.com/api/v1/voice/transcribe
Headers:  Authorization: Bearer abn_live_xxx
Form:     audio=<bytes>     # field name is "audio", not "file"

{
  "prompt": "warm acoustic 30-second cafe loop",
  "duration": 30
}

{
  "ok": true,
  "outputUrl": "https://...mp3",
  "creditsUsed": 90,
  "balance": 17867
}

{
  "imageUrl": "https://your-bucket/source.jpg",
  "scale": "4x"
}

{
  "ok": true,
  "outputUrl": "https://...png",
  "creditsUsed": 12,
  "balance": 17855
}

(no body)

{
  "ok": true,
  "models": [
    { "id": "google/nano-banana-pro/text-to-image", "type": "text-to-image", "credits": 43 },
    { "id": "openai/gpt-image-2/text-to-image", "type": "text-to-image", "credits": 18 },
    ...
  ]
}

{
  "modelId": "openai/gpt-image-2/text-to-image",
  "params": {
    "prompt": "A pixel-art mushroom",
    "aspect_ratio": "1:1",
    "quality": "medium",
    "resolution": "1k"
  }
}

{
  "ok": true,
  "outputUrl": "https://...png",
  "creditsUsed": 18,
  "balance": 17837
}

{
  "input": "Process new leads in HubSpot from today"
}

{
  "ok": true,
  "runId": "run_3a8b...",
  "status": "queued"
}

(no body)

{
  "ok": true,
  "balance": 18121
}

(no body)

{
  "ok": true,
  "service": "automatebusinessnow",
  "version": "v1",
  "timestamp": "2026-04-25T00:00:00.000Z"
}

Every endpoint has a per-key rate limit. The cap scales with your plan tier — Power users get 10× the Free baseline. When you exceed it, the response is 429 Rate limit exceeded and your credits are not charged.

Every error is shaped:

{
  "ok": false,
  "error": "human-readable sentence"
}

We don't ship a vendor SDK — every endpoint is plain HTTPS JSON, so any HTTP library works. We do publish two integration surfaces for AI tools:

• OpenAPI 3.1 spec at /api/v1/openapi— feed it to ChatGPT custom GPTs, Postman, or any tool-calling agent.
• MCP server config for Claude Desktop, Cursor, and other MCP-aware clients. Drop the snippet into your mcp.json and your AI assistant can call us directly. See /developers for the full config.