What you can do

Use BrowserAct from an MCP-compatible AI client to find and run published Bots or official templates, then retrieve their results. MCP (Model Context Protocol) is the connection that lets your AI client use BrowserAct tools.

Before you start

You need a BrowserAct account and an MCP-compatible client. Publish your own Bots before running them; official templates are ready to use.

Connect BrowserAct

BrowserAct MCP server URL:

https://mcp.browseract.com/mcp/

Prepare your general BrowserAct API Key and follow Connect to BrowserAct MCP to set up your AI client.

Run Bots or official templates

Describe your task to find a matching Bot or template. Your client checks the required inputs before running it.

Get your result

A run returns a task ID first. Your client uses that ID to check progress and retrieve the completed result and any available file information. Follow Run Bots and Templates for the full process.

Connect to BrowserAct MCP

Before you start

Create or copy a BrowserAct API Key on the API & Integrations page. For instructions, see API Keys & Authentication.

Connect with a setup prompt

Paste this prompt into your AI client to set up the connection. The server URL is included. Enter your API Key in the client configuration or secure credential field, never in chat.

Identify this MCP client and connect BrowserAct. Configure it if you have access; otherwise give the setup steps.

Name: browseract
URL: https://mcp.browseract.com/mcp/
Transport: Streamable HTTP
Authorization: Bearer 

Reuse a valid key, or get one at https://www.browseract.com/reception/integrations?co-from=link-Docs. Enter it in local configuration or a secure credential field, never in chat.

Reuse or repair an existing browseract connection; preserve other MCP settings. Verify that BrowserAct tools are available, and reload only if needed. Report any setup or verification issue. Do not run tasks.

Example confirmation after setup (wording varies by client):

Example client response confirming successful BrowserAct authentication

Copy a setup prompt for your client

Choose a client and click Copy. Paste API integration prompts into your coding assistant.

Client-specific setup prompts are available for Claude Desktop, Gemini CLI, Claude Code, Warp, Windsurf, VS Code, Cursor, Anthropic API, OpenAI API, OpenAI Agent Builder, and JSON configuration. Choose your client and copy its setup prompt.

Verify the connection

In your client’s MCP settings, confirm that browseract is enabled. The example uses Settings → Plugins → MCPs; paths vary by client.

BrowserAct enabled in the client MCP settings

Ask your client to confirm that search_bots and search_official_templates are available in this chat. If unavailable, see troubleshooting.

Try your first request

Send this request to your AI client:

Find a BrowserAct official template for my reddit post comment extraction task and show me its required inputs before running it.

Example response: a matching Reddit template and its required inputs. No extraction task has run yet.

Example response showing a Reddit template and its published inputs

Continue with Run Bots and Templates to submit a task and retrieve its result.

Run Bots and Templates

Run your published Bot

Describe what you want in natural language, or use a Bot ID as shown below. Your agent checks the required inputs and retrieves the results after a successful run.

Run BrowserAct Bot [BOT_ID]. Ask me for any missing required inputs, then run once and retrieve the results.

Run an official template

Describe what you want in natural language, or specify an official template ID. Your agent checks the required inputs and retrieves the results after a successful run.

Run BrowserAct official template [TEMPLATE_ID]. Ask me for any missing required inputs, then run once and retrieve the results.

Choose a proxy region for a template run

Ask your agent to find and use a supported region.

Find the BrowserAct proxy region code for [country or region] and use it when I ask you to run official template [TEMPLATE_ID].

Rate limits and performance

There is no request rate limit. The number of tasks you can run concurrently depends on your account’s current plan.

Available tools

Tool name	Category	Enabled by default	Description
`search_bots`	Bots	✅	Find published Bots accessible to your API Key.
`get_bot_details`	Bots	✅	Get Bot details and input schema.
`run_bot`	Bots	✅	Run a published Bot and return a task ID.
`search_official_templates`	Templates	✅	Find available official templates.
`get_official_template_details`	Templates	✅	Get template details and input schema.
`run_official_template`	Templates	✅	Run an official template and return a task ID.
`list_proxy_regions`	Regions	✅	List supported proxy region codes and names.
`list_tasks`	Tasks	✅	Find existing tasks and their IDs.
`get_task_status`	Tasks	✅	Get the current task status.
`get_task`	Tasks	✅	Get task details, output, and available files.
`cancel_task`	Tasks	✅	Cancel a running or paused task.

Troubleshooting and FAQ

Check the reported problem

Problem	What to check
Authentication fails	Use `Authorization: Bearer …`. The BrowserAct MCP connection requires a general API Key. Check that the key is still valid.
I see tools but not individual Bots	Bots and templates are discovered through search rather than listed individually in your client. Use `search_bots` to find your Bots or `search_official_templates` to find available official templates.
Input validation fails	Read the current resource details. Supply required fields with the correct types. Check parameter descriptions and defaults.
A run returned only an ID	This is expected for the BrowserAct MCP run tools. Query status, then retrieve the task result.
A run submission timed out	The request may already have created a task. If you have its ID, check that task. Otherwise, reconcile the submission using identifiable task information; do not submit again while the outcome is uncertain.
A region is rejected	Call `list_proxy_regions` and select a returned code.
A run fails	Read task details. Check the reported permission, input, balance, or website-access issue. Start another run only after you have addressed the issue and choose to run again.

Improve unclear input descriptions

If the client selects a tool or fills its inputs incorrectly, clarify the Bot and parameter descriptions.

Agent Built Bots

Open Edit Bot from the Bot details. Update the Bot description or the input descriptions under Parameters, then click Save.

Edit Agent Built Bot parameter descriptions and click Save

Workflow Built Bots

For Workflow Built Bots, open the input parameter editor in the Start node, update Description, and click OK to save. Then click Publish to make the updated configuration available through MCP. Ask your client to read the updated tool details.

Edit the Start node input description, save with OK, and publish the Bot

A request timed out

A timeout does not mean the task stopped. Query it by ID. If no ID was returned, check for the original task before retrying; do not create a duplicate while its status is uncertain.

Set up a custom MCP server with selected Bot tools

Use a Custom MCP Server to choose which published Bot tools this connection exposes. Both Agent Built and Workflow Built Bots can be selected.

1. Add a Server

Open Manage Custom MCP Servers. Click Add and complete the Server creation form.

2. Expose Bot tools

In the Server’s Action column, click the management icon shown below to open its settings.

Open management for the selected Custom MCP Server

Open Exposed Tools → Not Exposed, select the published Bot tools you want to make available, and click Expose. Complete any confirmation shown. Publishing a Bot does not automatically expose it to this Server.

Select Bot tools in Not Exposed and click Expose

3. Connect to your client

Open Connect to Clients. Copy the MCP Server URL and create or copy a key from the API Keys section on this Server’s connection page. Use both in your client’s configuration, following its setup guide.

Each Custom MCP Server requires its own independent MCP API Key. Use the URL and key from the same Server. Do not use your general BrowserAct API Key or a key from another Custom Server.

Connect to Clients shows the selected Server URL and its independent MCP API Keys

Keep the Server online and confirm that the selected tools appear in your client. Access remains subject to the key’s permissions.

Frequently asked questions

Do I need to expose each Bot before using the BrowserAct MCP server?

No. Connect with your BrowserAct API Key to search for accessible Bots and run your published Bots. You do not need to expose each Bot individually for this connection.

Do I need OAuth sign-in?

This connection uses API Key authentication. OAuth is not supported in this release.

Can I change which tools the BrowserAct MCP server exposes with a URL parameter?

No. URL ?tools= filtering is not supported.

Can I resume a task through MCP?

MCP does not provide a task-resume tool. Read the task details to decide your next step.