Resources · Claude connector

Claude TinyFish connectorWhat Claude can do in your TinyFish account.

The Claude TinyFish connector exposes 23 tools: 9 read, 6 write, 8 undocumented. Below: what Claude can read on the web, what the browser agent does in your name, what it costs, and the tools whose purpose nobody publishes.

Verified Trustpilot reviews · AI, automation & growth agency

Overview

Why hand your web chores to Claude and TinyFish

TinyFish gives Claude a cloud browser and a web agent. In practice you ask for a search, a heavy page to be read, or a run of clicks on a site, and Claude handles it through TinyFish. Per the official description, the agent runs in a remote Chrome browser and returns its results as structured JSON.

Search and read. search returns structured web results with titles, snippets and URLs. fetch_content pulls clean content from one to ten pages, including ones that need JavaScript to render.

Put an agent to work. run_web_automation carries out a multi-step task on a site from a URL and a plain-English goal: clicking, navigating, filling forms, logging in. Its sibling run_web_automation_async hands back control right away for long jobs, and get_run lets you check the outcome.

Watch the spend. get_wallet shows your TinyFish wallet balance and the rates that apply.

Now what the connector doesn't do. Topping up the wallet happens in the dashboard, not from Claude. The agent and the browser both draw on that balance. Eight tools on the sheet aren't described anywhere. And nothing starts on its own: Claude calls these tools while it answers you. A job that fires on a schedule is an automation tool's job, a different trade; the Integrations hub covers that side.

Vocabulary

The jargon, in one minute

Five terms you'll meet while connecting TinyFish.

Connector
The link you set up once between Claude and an account you already have, so Claude can work in it while it answers you.
Tool
One named thing a connector lets Claude do. Claude picks the ones it needs on its own, and the directory lists them by name.
Authorization
TinyFish's own sign-in screen, where you hand Claude the access it will use. Granted once per person and withdrawn the same way.
Approval
The confirmation Claude waits for before finishing something that changes your account, shown in the chat when it matters.
MCP
The shared standard connectors are built on, which is what lets an assistant like Claude talk to an outside service.
Connect

Connect TinyFish to Claude in three steps

  1. 01

    Find TinyFish in Claude

    In Claude's settings, open Customize, then Connectors, and look for TinyFish. On a Team or Enterprise workspace, an Owner or Primary Owner switches the connector on first so each member can sign in.

  2. 02

    Start the connection

    Hit Connect on its row. TinyFish opens its own sign-in page, where you confirm with Allow to link your TinyFish account. If the link breaks later, Disconnect and connect again.

  3. 03

    Read the authorization screen

    Read the authorization screen before you accept. It belongs to TinyFish, not Claude, and it sets what the access covers. You can also withdraw that access later from your TinyFish account.

Tools

All 23 tools, grouped by what they do

TinyFish gives Claude 23 tools: 9 that read your account, 6 that change something in it, and 8 no official source describes.

Three groups: what Claude reads, what it starts or changes, and what no official source describes. Tool names stay exactly as Claude shows them.

  • 9 read
  • 6 write
  • 8 not documented

What Claude reads (9)

9 tools

Nine tools that search, extract or check a status without launching anything.

batch_status

Checks where several runs stand at once, up to 8. TinyFish suggests polling every 30 to 60 seconds until all of them are done. For a single run, get_run gives the details.

When it helps
several runs are going at once and you want their status together.

Sourcedocs.tinyfish.ai · October 1, 2026 ↗

fetch_content

Pulls clean, structured content from one to ten pages, running JavaScript when needed, as Markdown, HTML or JSON. TinyFish recommends it over the agent when you only need to read.

When it helps
you only need to read a page, not interact with it.

Sourcedocs.tinyfish.ai · October 1, 2026 ↗

get_run

Brings back the details of one automation run from its ID: its status, its result and any error information.

When it helps
a background job should be done by now and you want to collect what it found, or understand why it failed.

Sourcedocs.tinyfish.ai · October 1, 2026 ↗

get_search_usage

Lists your past searches, filterable by date range and status, so you can see what was queried and when. The searches themselves aren't billed, per TinyFish's documentation.

When it helps
you want to find past searches by date or status.

Sourcedocs.tinyfish.ai · October 1, 2026 ↗

get_wallet

Shows your TinyFish wallet balance, auto-reload state, per-product rates and any top-up in progress. It only reads.

When it helps
before launching a big batch of automations, to make sure the balance will cover it.
Watch out
top-ups are handled in the dashboard, not here.

Sourcedocs.tinyfish.ai · October 1, 2026 ↗

list_browser_sessions

Lists all your remote browser sessions, with filters by session ID, time range and status, open or ended. Closing one early goes through close_browser_session, by its ID.

When it helps
you want to see which sessions are open and which have ended.

Sourcedocs.tinyfish.ai · October 1, 2026 ↗

list_fetch_usage

Gives a record of your past page fetches, filterable by date and status. It returns metadata only, never the page content itself. Fetching itself isn't billed, per TinyFish.

When it helps
you want to see which fetches ran, without the page content.

Sourcedocs.tinyfish.ai · October 1, 2026 ↗

list_runs

Lists your automation runs, with a status filter and pagination for long series. The details of one run come from get_run, and the status of several at once from batch_status.

When it helps
you want the runs that share a given status.

Sourcedocs.tinyfish.ai · October 1, 2026 ↗

What Claude starts or changes (6)

6 tools

Six tools that launch, stop or open something on TinyFish's side. No source spells out their confirmation, so the general rule applies.

batch_cancel

Approval: see the rule

Stops several running or pending runs in one go, up to 8, from their IDs. Running it again on runs that already finished changes nothing.

What Claude asks for
no source pins down a confirmation for bulk cancelling; see the general rule below.
When it helps
several runs are going and none of them is needed any more.

Sourcedocs.tinyfish.ai · October 1, 2026 ↗

cancel_run

Approval: see the rule

Halts one specific automation run, running or pending, by its ID. A run that has already ended simply reports its current status back.

What Claude asks for
nothing is documented on this for this tool, so the default rule described below holds.
When it helps
one run is no longer needed before it finishes.

Sourcedocs.tinyfish.ai · October 1, 2026 ↗

close_browser_session

Approval: see the rule

Shuts a browser session by its ID to free it straight away, instead of waiting for it to time out from inactivity.

What Claude asks for
TinyFish describes no specific prompt before closing; check the general approval rule.
When it helps
a session's work is done before its inactivity timeout.

Sourcedocs.tinyfish.ai · October 1, 2026 ↗

create_browser_session

Approval: see the rule

Starts a remote, stealth Chrome session and returns the connection details to drive it with Playwright, Puppeteer or Selenium.

What Claude asks for
no dedicated confirmation is described when a session opens; the shared rule applies.
When it helps
a developer wants to script a remote browser directly.
Watch out
per TinyFish, the browser draws from your wallet.

Sourcedocs.tinyfish.ai · October 1, 2026 ↗

run_web_automation

Approval: see the rule

Carries out a multi-step task on a site from a URL and a plain-English goal, streaming progress live: clicks, navigation, forms, logins.

What Claude asks for
no confirmation is documented for this specific launch; the approvals section sets the rule.
When it helps
a task needs clicks, navigation, a form or a login on a site.
Watch out
per TinyFish, the agent draws from your wallet.

Sourcedocs.tinyfish.ai · October 1, 2026 ↗

run_web_automation_async

Approval: see the rule

Starts the same automation as the web agent but immediately returns a run ID without waiting for it to finish. You check on the result separately.

What Claude asks for
the sources say nothing about a prompt before this deferred launch; the general rule covers it.
When it helps
a long job you don't need to watch step by step.

Sourcedocs.tinyfish.ai · October 1, 2026 ↗

Undocumented (8)

8 tools

The directory publishes these eight names and nothing more. TinyFish's connector docs don't describe them, and we won't guess their job from a name.

cancel_monitor

Only its name shows up on the directory's tool list. Neither TinyFish's connector documentation nor any Claude page describes it, so this page credits it with no capability of any kind.

Watch out
the general approval rule covers it whatever it does.

create_monitor

The directory lists this name, but no explanation comes with it, and TinyFish's tool documentation gives it no description, so this page doesn't describe it. No official text gives it a direction, read or write.

Watch out
whatever its effect, the general approval rule applies.

get_monitor

Nothing describes this tool beyond its name on the directory sheet. TinyFish leaves it out of its tool list, and Claude's documentation doesn't cover it either.

Watch out
the general approval rule, set out further down, still holds for it.

guide_next_step

This label is part of the directory sheet, and no official source documents it. TinyFish's documentation doesn't include it among its tools, so this page refrains from assigning it any role.

Watch out
it falls under the general approval rule.

list_monitors

The sheet publishes this name without a word of explanation, and no other official source fills the gap. This page won't venture a guess at its purpose.

Watch out
for this one too, it's the general approval rule that governs.

pause_monitor

No public documentation talks about this tool: the directory shows its name, and both TinyFish and Claude stay silent on how it works. The page therefore lends it no effect.

Watch out
the general approval rule covers it like the rest.

resume_monitor

All we know is the name written on the sheet. Neither TinyFish's connection guide nor any Anthropic source explains it, so this page says nothing about what it does or touches.

Watch out
the general approval rule applies to it.

run_monitor

Present on the directory list, undocumented by every official source: that's all that can be stated about it. The page credits it with neither reading nor writing, for lack of any text to back it up.

Watch out
it also falls under the general approval rule.
Approvals

What Claude checks with you first

By default, Claude stops and asks before every action it takes on an account for you. The request shows up in the chat at the moment it counts.

With TinyFish, six tools start or change something: launching the agent, cancelling, opening or closing a session. No source describes a confirmation specific to any of them, so the default rule covers them, as it does the eight undocumented tools. On Team and Enterprise workspaces, owners decide whether a member can let some actions through without being asked again, and can shut off writing for the whole org. Claude works with your TinyFish access and nothing beyond it.

Plans

Which plans include it

None of the 819 sheets in the official directory shows plan availability, and nobody publishes it connector by connector.

Anthropic's general rule is that remote connectors are open to all users on Claude, Cowork, Claude Desktop and mobile. On Team and Enterprise, an Owner or Primary Owner first switches the connector on for the org. On TinyFish's side you need an account, and both the agent and the browser draw from your wallet. For the current status, check the TinyFish sheet in the directory.

Limits

Where this connector stops

A connector isn't an automation. Claude calls these tools while it answers you, so nothing reruns by itself on a schedule.

Only TinyFish documents this connector, in its integration docs; neither Claude's help center nor Anthropic's docs have a page on it. Eight tools on the sheet remain undescribed. The official description also mentions monitoring websites for changes, but no source ties that to a specific tool on the sheet. The directory's partner badge isn't a security audit, and Anthropic notes on every sheet that it doesn't pick a publisher's tools or vouch for how they behave.

Need help

Need help connecting TinyFish to Claude?

A person reads every message.

FAQ

Claude TinyFish connector: common questions

01What can Claude do with the Claude TinyFish connector?
Claude can search the web, extract page content and put a browser agent to work on real websites, all from the chat. The documented tools cover structured search, fetching one to ten pages, launching multi-step automations, tracking and cancelling them, opening and closing browser sessions, usage history and your wallet balance. Eight more tools on the sheet have no source describing them, so this page doesn't assign them a role.
02Can Claude act on websites for me with TinyFish?
Yes. The web agent clicks, navigates, fills in forms and can sign in to a site, working from a plain-English goal. For signed-in tasks, TinyFish recommends a saved browser profile. Claude can also cancel runs and open or close browser sessions. Per TinyFish, the agent and the browser draw from your wallet, while search and fetch don't. Eight tools remain undocumented, so their effect isn't known, and they aren't counted among the writing tools.
03Does Claude ask before launching the TinyFish agent?
By default, Claude asks for confirmation before every action it takes on an account on your behalf. No source describes a confirmation specific to any TinyFish tool, so that default covers launching the agent, opening sessions and cancelling runs, as well as the undocumented tools. On Team and Enterprise, owners decide whether some members can skip those prompts, and can block writing for the whole organization. The official read-or-write label for each tool only shows in a connected account's settings.
04Which plans is the TinyFish connector available on?
No official source publishes plan availability for an individual connector, and no directory sheet displays it. Anthropic's general rule is that remote connectors are open to all users on Claude, Cowork, Claude Desktop and mobile, with an Owner or Primary Owner enabling them first on Team and Enterprise. On the publisher's side, you need a TinyFish account. For the current status, check the directory sheet while signed in. On Team and Enterprise, members can only connect once that activation is done.
05Does Claude see my whole TinyFish account?
No. Claude works with your rights and nothing more, so whatever your TinyFish access doesn't open stays closed to Claude too. The documented tools read your runs, sessions, usage history and balance, but topping up the wallet stays in the dashboard. In a company workspace, an owner can narrow what the connector is allowed to do for the whole organization, for example by keeping reading open while blocking agent runs.
06Why does the TinyFish authorization get stuck?
According to TinyFish, you need to be signed in to both claude.ai and your TinyFish account in your default browser before the authorization starts. That condition comes from TinyFish's own setup instructions, not from Anthropic. On TinyFish's side, the connection ends with a click on Allow on the publisher's sign-in page. If the link breaks later, Claude's rule is to disconnect the connector and connect it again from the list. The access can also be withdrawn from your TinyFish account.
07Claude or an automation tool for TinyFish?
They do different jobs, so pick by use case. Claude fits when you want to decide as the conversation goes what to search, which pages to read and when to launch the agent, approving along the way. An automation tool fits when a task has to run on its own on a schedule or after an event, because the connector never fires by itself and only acts while Claude is answering you.