Documentation
Bankworks renders web pages in a real Chrome browser and gives you the resulting HTML over HTTPS. You choose the URLs; we handle the browser, the network egress, block detection and retries.
Code samples
Every endpoint is shown in Python, Node.js and cURL. Pick a language on any example and the rest of the site remembers it.
- Python samples use
requests
(
pip install requests). - Node.js samples use the built-in
fetchand need Node 18 or newer. They are ES modules: save them as.mjs. - Every sample reads your key from the
API_KEYenvironment variable. Never paste a key into source you commit.
The model in one paragraph
You submit a job containing between 1 and 1,000 URLs.
Each URL becomes an independent task. Tasks are claimed
by workers, rendered, and stored. You poll the job for counters and
per-task status, then download each task's HTML. Nothing about this is
synchronous: POST /v1/jobs returns as soon as the work is
durably recorded, usually in well under a second.
Base URL and format
Every endpoint lives under /v1. Requests and responses are
JSON, except HTML downloads. Successful responses carry
"ok": true; failures carry "ok": false and a
human-readable error.
{"ok": false, "error": "Human-readable message"}
Endpoints at a glance
| Endpoint | Scope | Purpose |
|---|---|---|
POST /v1/jobs | jobs:create | Submit up to 1,000 URLs |
GET /v1/jobs/{jobId} | jobs:read | Counters and task list |
GET /v1/tasks/{taskId} | jobs:read | One task's metadata |
GET /v1/tasks/{taskId}/html | results:read | Download rendered HTML |
GET /v1/tasks/{taskId}/screenshot | results:read | Download the page screenshot |
GET /v1/me | any valid key | Who this key belongs to |
GET /v1/me/balance | any valid key | Your current balance |
GET /v1/me/audit | audit:read | Your own API activity |
POST /v1/ai-mode/tasks | ai-mode:create | Submit 1–100 business research tasks |
GET /v1/ai-mode/tasks/{taskId} | ai-mode:read | Poll task status |
GET /v1/ai-mode/tasks/{taskId}/result | ai-mode:read | Download the stored JSON result |
GET /healthz | none | Readiness probe |
Your responsibilities
You must have the right to collect the pages you request. We are infrastructure: we do not evaluate whether a target permits automated access, and submitting a URL is your assertion that you may collect it.
We respond to complaints from site operators by blocking the target for the account involved, and repeat offences end the account. Robots directives, terms of service and applicable law are yours to observe.
What the service does not do
- No standalone CAPTCHA-solving product. Supported CAPTCHA challenges may be handled only as part of an Advance Web Scraping API request, never past a login or an explicit refusal of access.
- Browsers are set up like an ordinary desktop browser (local clock for the exit, working graphics, no automation markers). This helps on sites that admit ordinary visitors; it is never used to get past a login or an explicit refusal of access.
- No residential proxy network built from consumer devices.
- No data extraction — you get HTML, and parsing is yours.
A site that still refuses us will keep refusing, and the API reports that honestly rather than storing a challenge page as a success.