Skip to content
Changelog

Report a bug

ZenRows integration: operations for AI agents

ZenRows connects to Type.com as an integration. This page lists the 44 operations documented for Claude and Codex in Type.com: 22 read, 15 write, 7 destructive.

About ZenRows

ZenRows is a web scraping API allowing developers to bypass CAPTCHAs and blocks, gather structured data from dynamic websites, and quickly integrate results into applications

Categories: Analytics & Data

How to connect ZenRows to Type.com

Access
Organization connection: one shared account for the workspace.

To add the connection, follow the steps in Space connections.

Documented operations

44 operations. Your connection's permissions determine which operations are available. API references describe the vendor surface; they do not expand your connection's credentials or access.

44 operations listed.

Fetch

  • Read

    Fetch a page

    Retrieve a URL through Zenrows and return its content. The same operation covers Extract (extract=auto or cssextractor), format conversion (responsetype), screenshots and structured output, because they are configurations of one request rather than separate endpoints.

  • Read

    Fetch a page with a POST

    Send a POST to the target through Zenrows. Takes the same query parameters as fetchPage; the request body is forwarded to the target as is, and the response carries the target's content.

  • Destructive

    Fetch a page with a PUT

    Send a PUT to the target through Zenrows. Takes the same query parameters as fetchPage; the request body is forwarded to the target as is, and the response carries the target's content.

Extract

  • Read

    List the domains Extract serves

    Every domain Extract can handle right now. A domain in this list is one you can send extract=auto to; a domain missing from it answers REQS007. Costs no credits. Documented at https://docs.zenrows.com/extract/endpoints.

  • Read

    List the domains your account submitted

    Every domain this account has submitted for preparation, with the monthly allowance alongside it. This lists what you submitted, not every domain Extract serves (see listExtractDomains). Costs no credits. Documented at https://docs.zenrows.com/extract/endpoints.

  • Read

    Read the monthly preparation allowance

    How many domains this account can still ask to have prepared this month. Costs no credits. Documented at https://docs.zenrows.com/extract/endpoints.

  • Read

    Track a domain's preparation

    How far a submission has got. Branch on status (preparing, ready, failed); state is the finer-grained step. Costs no credits. Documented at https://docs.zenrows.com/extract/endpoints.

  • Write

    Ask for a domain to be prepared for Extract

    Submit a domain for preparation, within a monthly allowance. The submission is accepted immediately and works through discovery, preparation and verification, typically in minutes: poll getExtractDomainStatus instead of blocking.

Onboarding

  • Read

    Get an API key for your agent, no signup required

    Start using Zenrows from your agent without signing up. This call returns a working API key straight away: no account setup, no login, no payment details. The key works right away on the free plan.

  • Read

    Check whether a person has claimed this account

    Report the claim state of an account created by createAgentAccount. Claiming does not rotate the key, so an agent keeps working across it; this is how it learns a human took ownership. Never contains an email or any secret.

Account

  • Read

    Read the plan behind the key

    Return the calling key's plan: credit allowance, credits used, and the concurrency it is allowed. Use it to budget before a large run.

Batch

  • Read

    List batch jobs

    Page through the jobs on the account, newest first. Keep paging while nextcursor is present.

  • Read

    Read a job

    The job and its latest run: progress in latestrun.stats, spend in latestrun.stats.spend. Poll with backoff until latestrun.status is terminal, or take a webhook. A deleted job answers 404.

  • Read

    Read a job's results

    Page through the per-URL results of the job's runs: status, a presigned resulturl valid for 2 hours, your externalid, and error on failed rows. Results are kept 14 days after the run finishes. A failed task is a row, not an HTTP error.

  • Read

    Download one task's content

    The stored content of one task, in the format the job requested. 409 while the task is still pending; 422 when it failed, with the stored error as the body.

  • Read

    List a job's runs

    Every run of the job, newest first, with its stats. Pages like listBatchJobs.

  • Read

    Export a run as a ZIP

    Ask for a bulk export of a run's content. It is prepared asynchronously: poll getBatchExport with the returned exportid until status is completed, then download downloadurl.

  • Read

    Check an export

    Read an export's state. Poll until status is completed (then download downloadurl, re-signed on each poll) or failed (then read error).

  • Read

    Read a job's webhook

    The webhook set on the job, if any.

  • Read

    List webhook signing keys

    Metadata of the HMAC keys your webhooks are signed with, never the secrets. An account with no key answers an empty object.

  • Write

    Create a batch job (many URLs in one request)

    Submit a job. A closed job carries all its URLs up front in tasks; an open job (status: open) is created empty and receives URLs through addBatchTasks until one call sets lastbatch. A scheduled job (type: scheduled) stores its URLs as a template and each fire creates a run.

  • Write

    Add URLs to an open job

    Append up to 10,000 URLs to an open job. Send lastbatch: true on the final call to close it; later additions then return 409. Can also return 409 for a few seconds while a very large submission is still being written: retry.

  • Write

    Close an open job

    Lock an open job so it accepts no further tasks. Equivalent to lastbatch on the last add.

  • Write

    Stop the current run

    Move the running or pending run to stopped, which is terminal. Tasks in flight may finish; pending ones stay as they are. Idempotent on a stopped run; 409 when the run already completed or was deleted.

  • Write

    Pause the current run

    Stop dispatching new tasks without ending the run; status does not change, pausestate does. Idempotent; 409 when the latest run is terminal. Not the same as pausing a schedule (setBatchScheduleState).

  • Write

    Resume a paused run

    Continue dispatching the remaining tasks of a paused run. Idempotent; 409 when the latest run is terminal.

  • Write

    Run a job again

    Start a new run of the same job: everything, or only what failed. Tasks that already succeeded are carried over, not fetched or charged again. Partial reruns are only allowed within 10 days of the original run finishing (409 partialrerunwindowexpired). Send no body.

  • Write

    Pause or resume a schedule

    Turn a scheduled job's future runs off or on without deleting its schedule. Separate from pausing a run (pauseBatchRun).

  • Write

    Upload a CSV of URLs

    Reserve an upload for a CSV too large to send inline (up to 50 MB and 100,000 rows; the slot lives 24 hours). Then PUT the raw file to upload.url with upload.headers, and pass fileinputid to createBatchJob.

  • Write

    Send a test webhook

    Deliver a sample notification to your endpoint so you can check the receiver and, with signature, your signature verification before a real run.

  • Write

    Start a key rotation

    Create the account's first signing key, or stage a candidate next to the active one so you can accept both while you deploy. The secret is returned once: store it.

  • Write

    Finish a key rotation

    Promote the candidate to active and retire the previous key.

  • Destructive

    Cancel the current run

    An alias of stopBatchRun.

  • Destructive

    Delete a run

    Remove a run's stored results. The job and its other runs are kept. Idempotent.

  • Destructive

    Set a job's schedule

    Replace a scheduled job's schedule. Give exactly one of at, rate or calendar, plus timezone for at and calendar. Times of day are full hours and runs are at least 60 minutes apart.

  • Destructive

    Set a job's webhook

    Be notified when a run ends (run.completed, run.failed) instead of polling. Deliveries are at least once: dedupe on the X-ZenRows-Event-Id header. With signature, each delivery carries X-Signature: t= ,v1= ,kid= , an HMAC-SHA256 of . with your key (see listHmacKeys).

  • Destructive

    Remove a job's webhook

    Stop delivering this job's notifications.

  • Destructive

    Cancel a key rotation

    Discard the staged candidate and keep signing with the active key.

Crawl

  • Read

    List your crawls

    Your crawls, newest first, without their results. Documented at https://docs.zenrows.com/crawl/endpointslist-your-crawls.

  • Read

    Read a crawl and its results

    The crawl, its coverage so far and one page of results. While it runs, nextcursor is never null: read again with the same cursor to get only the new URLs. Each URL is returned once. Documented at https://docs.zenrows.com/crawl/endpointsread-a-crawl.

  • Read

    Download all results

    Every result as one NDJSON file named {crawlid}.jsonl, one JSON object per line. Each line has the url; with outputformat, also contentstatus and, once fetched, the page HTML in content. While the crawl runs, a later download may hold more lines.

  • Read

    Read one page's HTML

    The HTML of one kept page, for a crawl started with outputformat. Take the path from the result's contenturl. Documented at https://docs.zenrows.com/crawl/endpointsread-a-page.

  • Write

    Start a crawl

    Start a crawl from one URL and return at once with 202 Accepted, the crawl in the body and its path in the Location header. Each page it opens is one Fetch request with Adaptive Stealth Mode, billed 1 to 25 credits; failed pages are not billed. Documented at https://docs.zenrows.

  • Write

    Stop a crawl

    Stop a running crawl. No new page is fetched; pages already being fetched finish, are billed and are kept for up to 10 minutes. Read the crawl until nextcursor is null for its final results. A crawl that already ended returns its final state. Send no body.

See How connections work for access and permissions, or browse all integrations.