API Documentation

Authentication

All API requests require authentication via an API key. Include your key in the X-API-Key header.

X-API-Key: ao_live_your_api_key_here

Alternatively, pass the API key in the request body:

{
  "apiKey": "ao_live_your_api_key_here",
  "company": "Stripe"
}

Run Agent

Start an agent run. The job is queued and processed asynchronously. Use the returned job_id to check status.

curl -X POST https://applyoperator.com/api/v1/agents/company-intel/run \
  -H "X-API-Key: ao_live_your_key" \
  -H "Content-Type: application/json" \
  -d '{
    "company": "Monzo",
    "domain": "monzo.com",
    "depth": "standard"
  }'

import requests

response = requests.post(
    "https://applyoperator.com/api/v1/agents/company-intel/run",
    headers={"X-API-Key": "ao_live_your_key"},
    json={
        "company": "Monzo",
        "domain": "monzo.com",
        "depth": "standard"
    }
)

data = response.json()
print(f"Job ID: {data['job_id']}")

const response = await fetch(
  "https://applyoperator.com/api/v1/agents/company-intel/run",
  {
    method: "POST",
    headers: {
      "X-API-Key": "ao_live_your_key",
      "Content-Type": "application/json"
    },
    body: JSON.stringify({
      company: "Monzo",
      domain: "monzo.com",
      depth: "standard"
    })
  }
);

const data = await response.json();
console.log("Job ID:", data.job_id);

Response

{
  "success": true,
  "job_id": "ci_abc123xyz",
  "status": "queued",
  "estimated_time": "30-120 seconds"
}

Jobs

Check job status and retrieve results.

{ "job_id": "ci_abc123xyz" }

{
  "success": true,
  "job": {
    "id": "ci_abc123xyz",
    "agent": "company-intel",
    "status": "completed",
    "progress": 100,
    "results": { "..." },
    "cost": { "amount": 1.00, "currency": "USD" },
    "created_at": "2026-01-15T10:30:00Z",
    "completed_at": "2026-01-15T10:31:23Z"
  }
}

Webhooks

Register a webhook URL to receive notifications when jobs complete or fail.

{
  "url": "https://your-app.com/webhook",
  "events": ["job.completed", "job.failed"]
}

Account

Tool Code

Every marketplace tool exports an async function handler(input, api). The platform calls this function inside a sandboxed Linux namespace with no direct network access.

Handler Signature

async function handler(input, api) {
  // input  — JSON object sent by the caller
  // api    — sandbox API: fetch, log, memory, storage
  // return — any JSON-serializable value (sent back as job output)

  api.log('Starting agent run');

  const res = await api.fetch('https://api.example.com/data', {
    method: 'POST',
    headers: { 'Authorization': 'Bearer ...' },
    body: JSON.stringify({ query: input.query })
  });

  const prev = await api.memory.read();
  await api.memory.write(prev + '\n' + new Date().toISOString());

  return { status: res.status, data: res.body };
}

Sandbox API

The api object provides controlled access to the outside world from inside the sandbox.

api.fetch(url, options)

Make HTTP requests to external services. All requests are proxied through a bridge socket — the sandbox has no direct network.

Returns: { status, statusText, headers, body, text }

Restrictions: HTTPS only, no private/localhost IPs, 15s timeout, 1 MB max response size. body is auto-parsed as JSON when possible, text is always the raw string.

api.log(message)

Capture log messages visible in job results. Messages are truncated to 10,000 characters. Logs are returned alongside the agent output when the job completes.

api.memory

Persistent markdown-based memory stored in SQLite. Two scopes are available:

Memory persists across runs. Agent memory is scoped to the specific agent slug; global memory is shared across all agents for the same user.

api.storage

Paid disk storage for files. Requires an active storage subscription (see Dashboard → Storage).

api.getCredentials()

Agent tools often need user-specific secrets such as login emails, passwords, or API keys. Instead of hardcoding these values, users save credentials once in the Agent Tools page. Saved credentials are automatically injected at runtime.

In your tool's input_schema, add "credential": true to any property that contains sensitive data. The platform uses this flag to identify which fields should be saved separately and never sent to the LLM.

{
  "type": "object",
  "properties": {
    "email": { "type": "string", "description": "Login email", "credential": true },
    "password": { "type": "string", "description": "Login password", "credential": true },
    "date_range": { "type": "string", "description": "Date range to download" }
  }
}

Fields marked credential: true appear in the Agent Tools configuration form and can be saved by the user. Non-credential fields (like date_range above) are passed as regular input each time the tool runs.

Returns: { configured, credentials }

async function handler(input, api) {
  const creds = await api.getCredentials();
  if (!creds.configured) {
    return { error: 'Credentials not configured. Save them in the Agent Tools page.' };
  }

  const { email, password } = creds.credentials;

  // Use credentials to authenticate
  const res = await api.fetch('https://example.com/api/login', {
    method: 'POST',
    body: JSON.stringify({ email, password })
  });

  return { status: 'ok', data: res.body };
}

Users save credentials in Dashboard → Agent Tools. Values are stored per-user, scoped to the tool slug. At runtime, api.getCredentials() returns the actual values inside the sandbox. Credential fields are also auto-injected into the tool's input when called by the operator.

Sandbox Limits

Agents run inside isolated Linux namespaces with the following resource constraints:

Telegram Integration

Connect a Telegram bot to receive notifications and chat with your AI operator agents directly from Telegram.

Prerequisites

• A Telegram account
• A Telegram bot created via @BotFather

Setup Steps

Supported Notifications

• Job completed / failed
• Low credits warning / credits depleted
• Agent marketplace approval

Configure which notifications are sent via Telegram in Notification Preferences.

Bidirectional Chat

When sessions are linked, you can send messages to the bot and the operator agent will process them and respond. The agent has full access to its configured tools (web search, email, etc.).

Multi-Session Chat

You can link multiple agent sessions to your Telegram bot and switch between them using commands:

All regular messages (not starting with /) are sent to the currently active session. Switch between agents at any time.

WhatsApp Integration

Connect WhatsApp to receive notifications and chat with your AI operator agents directly from your phone.

Prerequisites

• A WhatsApp account with an active phone number
• WhatsApp installed on your phone

Setup Steps

Supported Notifications

• Job completed / failed
• Low credits warning / credits depleted
• Agent marketplace approval

Configure which notifications are sent via WhatsApp in Notification Preferences.

Multi-Session Chat

You can link multiple agent sessions and switch between them using commands:

All regular messages are sent to the currently active session. Switch between agents at any time.

Troubleshooting

Storage Sync (WebDAV)

Sync your agent storage files between your local computer and ApplyOperator using the WebDAV protocol. Mount your storage as a native folder, use bidirectional sync, or automate backups — works on Windows, Linux, and macOS.

Prerequisites

• An active storage subscription (see Dashboard → Storage)
• An API key from Settings

Authentication

The WebDAV endpoint uses HTTP Basic Auth. Set your username to anything (e.g. your email) and use your API key as the password:

URL:      https://applyoperator.com/webdav/
Username: your-email@example.com   (any value)
Password: ao_live_your_api_key_here

Quick Start with rclone

rclone is a free, open-source tool for syncing files with cloud storage. It supports WebDAV and runs on all major operating systems.

Type:     webdav
URL:      https://applyoperator.com/webdav/
Vendor:   other
User:     your-email@example.com
Password: ao_live_your_api_key_here

# Download remote files to local folder
rclone sync ao: ~/ao-storage/

# Upload local files to remote
rclone sync ~/ao-storage/ ao:

# Bidirectional sync (keeps both sides up to date)
rclone bisync ~/ao-storage/ ao:

# List remote files
rclone ls ao:

# Mount as a native folder (files appear like local files)
rclone mount ao: ~/ao-storage/ --vfs-cache-mode full

# The mount stays active until you press Ctrl+C
# On Windows, use a drive letter: rclone mount ao: X: --vfs-cache-mode full

Native OS Mounting

You can also connect directly from your OS file manager without installing rclone:

curl Examples

You can also interact with the WebDAV endpoint directly using curl:

# List files (PROPFIND)
curl -X PROPFIND https://applyoperator.com/webdav/ \
  -u "user:ao_live_your_key" \
  -H "Depth: 1"

# Create a directory
curl -X MKCOL https://applyoperator.com/webdav/my-agent/ \
  -u "user:ao_live_your_key"

# Upload a file
curl -X PUT https://applyoperator.com/webdav/my-agent/data.json \
  -u "user:ao_live_your_key" \
  -d '{"key": "value"}'

# Download a file
curl https://applyoperator.com/webdav/my-agent/data.json \
  -u "user:ao_live_your_key"

# Delete a file
curl -X DELETE https://applyoperator.com/webdav/my-agent/data.json \
  -u "user:ao_live_your_key"

Access Control