Oassis API documentation

Give an agent access to pages, whole sites and a persistent browser. Choose a tool for each step and keep the result in a form your application can use.

Three API families

Web Tools — /web/v1: individual operations. Web Tasks — /tasks/v1: one-shot extraction and research jobs. Agents — /agents/v1: scheduled monitoring services.

Extraction: POST /tasks/v1/scrape/batch or POST /tasks/v1/crawl. Add /:id for GET status and /:id/cancel for POST cancellation. Research uses /tasks/v1/research with the same lifecycle. Cancel an agent registration with POST /agents/v1/services/:id/cancel; the last report is retained.

Choose your level of control

Web Tools: choose an individual search, page read, map or browser action. Web Tasks: run batch extraction, crawling or research as an asynchronous job. Read the Web Tasks documentation.

Choose your connection

Connect an MCP client to https://api.oassis.dev/mcp to discover tools. MCP includes 100 trial credits for scrape and map. Create an API key to use all tools with your shared balance; verified accounts receive 1,000 welcome credits. HTTP accepts your API key and shared credits, or x402: first read the price challenge, then pay and retry.

Web Tools

MCP tool	HTTP route	What it does
`web_scrape`	`POST /web/v1/scrape`	Read a page or document in the requested formats.
`web_map`	`POST /web/v1/map`	List a site's URLs from its sitemap and optionally the page.
`web_search_exa`	`POST /web/v1/search`	Search the web.
`web_session_open`	`POST /web/v1/session`	Open a browser session and inspect its controls.
`web_act`	`POST /web/v1/act`	Interact with an open browser session.
`web_session_close`	`POST /web/v1/session/{id}/close`	Close the browser session and settle any outstanding time.
`web_feedback`	`POST /web/v1/feedback`	Report whether a result was correct. Free.

Account and credits are separate: GET /web/v1/account requires your API key.

Web Tasks

MCP tool	HTTP route	What it does
`web_scrape_batch`	`POST /tasks/v1/scrape/batch`	Start a batch extraction job from a supplied list of URLs.
`web_batch_status`	`GET /tasks/v1/scrape/batch/{id}`	Read batch status and collect its results.
`web_batch_status`	`POST /tasks/v1/scrape/batch/{id}/cancel`	Cancel pending batch work. With MCP, send cancel: true.
`web_crawl`	`POST /tasks/v1/crawl`	Start a crawl job that discovers and reads linked pages.
`web_crawl_status`	`GET /tasks/v1/crawl/{id}`	Read crawl status and collect its results.
`web_crawl_status`	`POST /tasks/v1/crawl/{id}/cancel`	Cancel pending crawl work. With MCP, send cancel: true.
`web_task_start`	`POST /tasks/v1/research`	Start research from public pages and return an answer with sources.
`web_task_status`	`GET /tasks/v1/research/{id}`	Read research progress, sources, result and billing.
`web_task_cancel`	`POST /tasks/v1/research/{id}/cancel`	Cancel pending research work.

Use web_task_start, web_task_status and web_task_cancel to delegate and follow research. See the full Tasks guide, HTTP requests, MCP examples and pricing.

Credits and API keys

1 credit = $0.002; a basic page uses 1 credit. Buy 2,500 / 10,000 / 25,000 credits for $5 / $20 / $50, once or monthly. Purchased credits accumulate without scheduled expiry. Create a verified account for 1,000 welcome credits once and generate an API key.

curl -i https://api.oassis.dev/web/v1/scrape -H 'Authorization: Bearer YOUR_API_KEY' -H 'Content-Type: application/json' -d '{"url":"https://example.com","formats":["markdown"]}'

Use the same Authorization header in your MCP client. Check available and reserved credits with GET /web/v1/account.

Read your first page

This HTTP request asks for Markdown. With no payment, expect a 402 response describing the required payment; plain curl does not sign or pay it.

curl -i https://api.oassis.dev/web/v1/scrape \
  -H 'Content-Type: application/json' \
  -d '{"url":"https://example.com","formats":["markdown"]}'

Build a web workflow

Use web_search_exa when you need to discover relevant pages. Search requires credits with an API key, or an HTTP x402 payment.
Use web_scrape for known URLs and web_map to discover links on a site.
Use web_crawl to read multiple pages within an explicit limit.
Use web_session_open and web_act when the page requires interaction.

Inspect errors before continuing

HTTP 402 means payment is required; 400 means invalid arguments. MCP may return a tool error inside an HTTP 200 response, so inspect the JSON-RPC result. Batch and crawl operations return a job identifier: collect their results through the status tool or route rather than repeating the start request.

Continue with MCP integration, x402 payments or the OpenAPI reference.