ceven

TinyFish MCP

Controls a headless browser to scrape data, click buttons, and fill forms on any website that lacks an API, then pipes that live web data directly into your workflows.

Try TinyFish MCP in Ceven

Ask Ceven anything
Standard

Why use Ceven?

  1. AI native TinyFish MCP integration

    • Describe the outcome and Ceven picks the right TinyFish MCP calls, fills the parameters, and checks the result.
    • Structured, agent friendly tool schemas so each call runs reliably instead of by guesswork.
    • Rich coverage for reading, writing, and querying your TinyFish MCP data, across all 8 of its actions.

  2. Managed auth

    • Built in OAuth with automatic token refresh and rotation.
    • One place to manage, scope, and revoke TinyFish MCP access.
    • Per user and per environment credentials instead of shared keys.

  3. Agent optimized design

    • Actions are tuned from real success and error rates so reliability climbs over time.
    • Full execution logs so you always know what ran in TinyFish MCP, when, and on whose behalf.
    • The agent pauses and asks when TinyFish MCP is unclear instead of plowing ahead.

  4. Enterprise grade security

    • Fine grained access so you control which agents and people can reach TinyFish MCP.
    • Least privilege by default, read scopes first and only the writes a workflow needs.
    • A full audit trail of every TinyFish MCP action to support review and sign off.

Supported tools

Every action Ceven's agents can run on TinyFish MCP, and when to use it.

Navigate to URL
Use this when the workflow needs to load a specific web page to begin a task or extract information.
Click element
Trigger a click on a button, link, or menu item using a CSS selector or text label.
Type text
Enter text into a search bar, login field, or form input to interact with the site.
Get page content
Pull the full text or HTML of the current page to analyze the content or find data points.
Take screenshot
Capture a visual image of the current browser state to verify layout or save a record.
Wait for selector
Pause the workflow until a specific element appears on the page, useful for slow loading sites.
Execute JavaScript
Run a custom script in the browser console to manipulate the page or extract hidden data.
Get page title
Pull the title of the current tab to confirm the agent is on the correct page.
Press key
Simulate a keyboard press like Enter or Escape to submit forms or close modals.
Scroll page
Move the page view down or up to trigger lazy loading of content or reach the footer.
Hover over element
Simulate a mouse hover to reveal hidden tooltips or dropdown menus.
Clear cookies
Reset the browser session to start a fresh visit without cached data or login states.

12 actions · scroll to see them all

Frequently asked questions

TinyFish MCP attempts to navigate around basic bot detection using human like interaction patterns and randomized delays. However, complex CAPTCHAs like reCAPTCHA v3 or Cloudflare Turnstile can sometimes block the agent. In these cases, the workflow will return an error indicating a challenge was encountered. You can handle this by adding a step to notify a human operator to solve the puzzle manually or by using a third party CAPTCHA solving service through a separate API call in your Ceven workflow before returning to the TinyFish browser session.
Yes, TinyFish MCP supports session persistence through cookie management. When you perform a login action, the browser stores the session cookies. You can configure your Ceven workflow to save these cookies to a secure vault and inject them back into the browser instance at the start of the next run. This prevents the agent from having to perform the full login flow every time, which reduces the risk of triggering security alerts or getting locked out for too many login attempts in a short window.
TinyFish MCP enforces a concurrency limit based on your subscription tier. For most users, this means you can run up to five browser sessions simultaneously. If you attempt to launch more, the API will return a rate limit error. To avoid this, we recommend queuing your web tasks in Ceven rather than firing them all at once. Be aware that some websites also have their own rate limits and may temporarily ban your proxy IP if the agent requests too many pages per second.
Absolutely. Unlike simple HTML scrapers that only see the initial page load, TinyFish MCP uses a full headless browser. This means it waits for JavaScript to execute and the DOM to fully render before attempting to read data or click elements. If a page uses a framework like React or Vue, the agent interacts with the rendered output exactly as a user would. You can use the Wait for selector action to ensure the dynamic content has loaded before the agent tries to scrape it.
The agent uses a combination of CSS selectors, XPaths, and text matching. When you tell the agent to click a button, it searches the page source for the most likely match. For example, if you say click the Submit button, it looks for an element with the text Submit or an ID containing the word submit. For highly complex pages, you can provide the exact CSS selector in the workflow configuration to ensure the agent never clicks the wrong element.
Yes, the agent can interact with file input fields to upload documents from a URL or a base64 string provided by your workflow. For downloads, TinyFish captures the file stream and can save it to a specified cloud storage location like S3 or Google Drive. You just need to ensure the workflow has the correct permissions to write to that destination. Note that some sites require a specific browser profile to handle downloads without a prompt, which TinyFish manages automatically.
All browser sessions are isolated in ephemeral containers that are destroyed immediately after the workflow completes. TinyFish does not store your browsing history or the content of the pages you visit on their own servers. Any data extracted is passed directly back to Ceven over an encrypted connection. If you are handling sensitive credentials, we recommend using the cookie injection method rather than typing passwords in plain text within your workflow logs to maintain a high security posture.
If a site changes its HTML structure, a hard coded CSS selector might fail. However, because Ceven uses an LLM to drive TinyFish MCP, the agent can often self heal. If the primary selector fails, the agent can scan the page for similar text or labels to find the new location of the element. If the change is too drastic, the workflow will trigger a failure notification, allowing you to update the selector once and resume the automation across all your active agents.

Alternatives to TinyFish MCP

Other tools that solve a similar problem. Ceven supports these too, so you can switch or run more than one at once.

Browse AI logoBrowse AIApify logoApifybrowserless logobrowserless

Try Ceven on your stack

Plug Ceven on top of the tools you already run. Connect TinyFish MCP and the rest of your stack, describe the outcome, and its agents handle the work end to end, days of it in minutes.

Get started for free