---
title: "Use Claude Code with Browser automation — tool setup and examples — AgentSky"
description: "Use Claude Code with the Browser automation tool on AgentSky. Preselect the capability, run a bounded task, and verify the returned result and usage."
url: https://agentsky.dev/agents/claude-code/tools/browser-automation
---

# Claude Code + Browser automation

Claude Code runs the work loop and Browser automation performs the selected operation. The Composer below preselects Claude Code, Claude Opus 5.5, and this tool so you can test one real task with the same billing and recovery path as any AgentSky run.

[Repo](https://agentsky.dev/auth/start?returnTo=%2Fagents%2Fclaude-code%2Ftools%2Fbrowser-automation%3Fresume%3D1)

How the combination works

## Claude Code decides when to call Browser automation.

Claude Code runs the coding loop, reads and edits repository files, invokes tools, and reports the result from its cloud workspace.

Use TinyFish Agent when the task requires clicks, typing, navigation or structured extraction rather than a read-only fetch or CDP session. This capability is classified as a public action and is never added to a new agent by default.

- Claude Code Owns the task loop, context, model calls, tool choice, cloud computer, progress and final result.
- Browser automation Performs the operation exposed by the current Browser automation capability. Its input limits and output shape come from the live tool guide and adapter.
- AgentSky Keeps the selected agent, model, capability grant, task state and usage record together. Tool charges and model or runtime charges remain itemized separately.

Setup and first task

## Start with one bounded browser automation task.

The Composer carries Claude Code, Claude Opus 5.5, and Browser automation. Begin with the smallest input that lets you inspect the output before expanding the workflow.

01

### Open the Claude Code Composer

The Composer above preselects Claude Code, Claude Opus 5.5, and Browser automation. Add the target, required inputs and a concrete expected result; the draft stays in place through sign-in.

02

### State the Browser automation boundary

Open the public directory, reveal all visible rows and return names, descriptions and source links. Stop without submitting any form.

03

### Run and verify the returned result

Check the output against the source or brief, confirm that the result has the expected format, and inspect the task usage for model, runtime and tool charges.

Inputs, cost and limits

## What Claude Code + Browser automation establishes

Repository work where Claude Code's coding workflow and supported Claude models fit the team's existing review process.

Configure and verify

- Claude Code and Claude Opus 5.5 are preselected for the browser evaluation.
- Browser automation is enabled from the current capability catalog with its live input and output guidance.
- The first task can be checked from the returned artifact, text or provider result before a larger workflow is attempted.

Still depends on the task

- A tool guide does not guarantee a provider result for malformed, blocked or unsupported Browser automation inputs.
- A successful tool call does not establish that the result is correct for your business decision; inspect the output at its source boundary.
- Total task cost can include model tokens, active computer time and the selected tool's billable units.

Current Browser automation rates

### tinyfish.agent

$0.032 per step.

Tool-specific details

## Use the current Browser automation contract.

Use TinyFish Agent when the task requires clicks, typing, navigation or structured extraction rather than a read-only fetch or CDP session. This capability is classified as a public action and is never added to a new agent by default.

- Explicit grant tinyfish.agent is separate from tinyfish.browser. Existing browser-session grants do not authorize automated clicks, form input or submissions.
- Input and result Provide a public URL and goal. The legacy task field is accepted as an alias. A completed call returns the structured result, run id, terminal status and billable step count.
- Bounded execution The request caps TinyFish Agent at 50 steps. Strict mode, screenshots and snapshots are not available.

## Claude Code + Browser automation questions

## Related resources

[Claude CodeReview Claude Code's supported models, tools, channels and connectors.](https://agentsky.dev/agents/claude-code) [Browser automation toolReview the current operation inputs, outputs, examples and rates.](https://agentsky.dev/features/tools/browser-automation) [Claude Code + Claude Opus 5.5Review the current default agent and model combination.](https://agentsky.dev/agents/claude-code/with/claude-opus-5-5) [PricingReview model, active-computer and tool charges.](https://agentsky.dev/pricing)

## Give Claude Code a browser automation task.

[Open the Composer](https://agentsky.dev/agents/claude-code/tools/browser-automation#composer)

## Frequently asked questions

### Can Claude Code use Browser automation on AgentSky?

Yes. This page is generated from the current AgentSky tool catalog and preselects Browser automation for Claude Code. The exact operation still depends on the tool's current adapter, required inputs and provider availability.

### What does Claude Code do with Browser automation?

Claude Code interprets the task, decides whether the permitted Browser automation operation is useful, supplies its inputs, and returns the result in the running task. The tool performs the operation; it does not replace the agent's reasoning or verification.

### What should my first Claude Code + Browser automation task be?

Start with one small, read-only or reversible request. Open the public directory, reveal all visible rows and return names, descriptions and source links. Stop without submitting any form. Then compare the returned result with the source or brief before adding more inputs or allowing follow-up actions.

### How do I verify a Browser automation result?

Inspect the returned artifact or response at its source boundary. Check the expected format, important fields, fidelity to the brief and any provider error or partial-result signal; a completed tool call alone does not prove correctness.

### How much does Claude Code + Browser automation cost?

There is no fixed task price. AgentSky records the selected tool's billable units together with model token usage and active cloud-computer time. Review the current Browser automation guide for the operation rate; provider charges or quotas can remain separate.

### What happens if I am signed out or my balance is empty?

The shared Composer preserves the task, Claude Code, Claude Opus 5.5, and Browser automation through sign-in. Before dispatch, AgentSky checks the account and balance; if funding is required, it shows the billing path and keeps the draft instead of silently changing the selections.

### Can I use Browser automation without Claude Code?

You can call the underlying tool API directly when you need an operation without a cloud agent. This page is for the combined workflow, where Claude Code supplies the task loop and Browser automation is available as a selected capability in the Composer.

### Can I combine Browser automation with a connector?

Yes, when the task needs both a tool operation and authorized application data. Enable the connector separately, state the resource and permission boundary, and verify each result at its own source; tool access does not grant connector access.

---

Markdown copy of https://agentsky.dev/agents/claude-code/tools/browser-automation. Every agentsky.dev page has one: add `.md` to its address or send `Accept: text/markdown`. Start from https://agentsky.dev/llms.txt.
