Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@Browserbase MCP Server Navigate to example.com and extract the page title."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
The Model Context Protocol (MCP) is an open protocol that enables seamless integration between LLM applications and external data sources and tools. Whether you're building an AI-powered IDE, enhancing a chat interface, or creating custom AI workflows, MCP provides a standardized way to connect LLMs with the context they need.
This server provides cloud browser automation capabilities using Browserbase and Stagehand. It enables LLMs to interact with web pages, extract information, and perform automated actions.
This is a self-hostable version of the Browserbase hosted MCP server with the same tools and functionality. We recommend using the hosted version for the easiest setup.
This server exposes 6 tools that match the hosted Browserbase MCP server:
Tool | Description | Input |
| Create or reuse a Browserbase session | (none) |
| Close the current Browserbase session | (none) |
| Navigate to a URL |
|
| Perform an action on the page |
|
| Observe actionable elements on the page |
|
| Extract data from the page |
|
Related MCP server: Browserbase MCP Server
We currently support 2 transports for our MCP server, STDIO and SHTTP. We recommend you use SHTTP with our hosted MCP server to take advantage of the server at full capacity.
Use the Browserbase hosted MCP server at https://mcp.browserbase.com/mcp. This is the easiest way to get started -- we host the server and provide the LLM costs for Gemini, the best performing model in Stagehand.
For full setup instructions, see the Browserbase MCP documentation.
If your client supports SHTTP:
{
"mcpServers": {
"browserbase": {
"type": "http",
"url": "https://mcp.browserbase.com/mcp"
}
}
}If your client doesn't support SHTTP:
{
"mcpServers": {
"browserbase": {
"command": "npx",
"args": ["mcp-remote", "https://mcp.browserbase.com/mcp"]
}
}
}You can either use our server hosted on NPM or run it completely locally by cloning this repo.
Note: If you want to use a different model you have to add --modelName to the args and provide that respective key as an arg. More info below.
Go into your MCP Config JSON and add the Browserbase Server:
{
"mcpServers": {
"browserbase": {
"command": "npx",
"args": ["@browserbasehq/mcp"],
"env": {
"BROWSERBASE_API_KEY": "",
"BROWSERBASE_PROJECT_ID": "",
"GEMINI_API_KEY": ""
}
}
}
}That's it! Reload your MCP client and you're ready to go.
git clone https://github.com/browserbase/mcp-server-browserbase.git
cd mcp-server-browserbase
npm install && npm run buildgit clone https://github.com/browserbase/mcp-server-browserbase.git
cd mcp-server-browserbase
docker build -t mcp-browserbase .Then in your MCP Config JSON run the server:
{
"mcpServers": {
"browserbase": {
"command": "node",
"args": ["/path/to/mcp-server-browserbase/cli.js"],
"env": {
"BROWSERBASE_API_KEY": "",
"BROWSERBASE_PROJECT_ID": "",
"GEMINI_API_KEY": ""
}
}
}
}{
"mcpServers": {
"browserbase": {
"command": "docker",
"args": [
"run",
"--rm",
"-i",
"-e",
"BROWSERBASE_API_KEY",
"-e",
"BROWSERBASE_PROJECT_ID",
"-e",
"GEMINI_API_KEY",
"mcp-browserbase"
],
"env": {
"BROWSERBASE_API_KEY": "",
"BROWSERBASE_PROJECT_ID": "",
"GEMINI_API_KEY": ""
}
}
}
}The Browserbase MCP server accepts the following command-line flags:
Flag | Description |
| Enable Browserbase proxies for the session |
| Enable Browserbase Verified Identity (Only for Scale Plan Users) |
| Deprecated alias for |
| Enable Browserbase Keep Alive Session |
| Specify a Browserbase Context ID to use |
| Whether to persist the Browserbase context (default: true) |
| Port to listen on for HTTP/SHTTP transport |
| Host to bind server to (default: localhost, use 0.0.0.0 for all interfaces) |
| Browser viewport width (default: 1024) |
| Browser viewport height (default: 768) |
| The model to use for Stagehand (default: google/gemini-2.5-flash-lite) |
| API key for the custom model provider (required when using custom models) |
| Enable experimental features (default: false) |
These flags can be passed directly to the CLI or configured in your MCP configuration file.
Note: These flags can only be used with the self-hosted server (npx @browserbasehq/mcp or Docker).
Stagehand defaults to using Google's Gemini 2.5 Flash Lite model, but you can configure it to use other models like GPT-4o, Claude, or other providers.
Important: When using any custom model (non-default), you must provide your own API key for that model provider using the --modelApiKey flag.
{
"mcpServers": {
"browserbase": {
"command": "npx",
"args": [
"@browserbasehq/mcp",
"--modelName",
"anthropic/claude-sonnet-4.5",
"--modelApiKey",
"your-anthropic-api-key"
],
"env": {
"BROWSERBASE_API_KEY": "",
"BROWSERBASE_PROJECT_ID": ""
}
}
}
}Note: The model must be supported in Stagehand. Check out the docs here.
Licensed under the Apache 2.0 License.
Copyright 2025 Browserbase, Inc.
Perform an action on the page
| Name | Required | Description | Default |
|---|---|---|---|
| action | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations, the description must disclose behavioral traits, but it fails to mention any effects (e.g., whether actions modify the page, require a loaded page, or have side effects). The agent has no insight into tool behavior.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is very concise (one sentence), but it sacrifices informativeness. While it is not verbose, it lacks structure and essential details, making it insufficient.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool's simplicity (one parameter, no output schema, no annotations), the description is critically incomplete. An agent cannot determine valid inputs, expected outcomes, or when to invoke this tool.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The single parameter 'action' is a plain string with no description in schema or tool description. The schema coverage is 0%, and the description adds no meaning beyond the parameter name. No examples, enums, or constraints are provided.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description 'Perform an action on the page' is vague and does not specify what kinds of actions are supported (e.g., click, type, scroll). It provides minimal differentiation from sibling tools like 'navigate' or 'observe', which also involve actions on the page.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
No guidance is given on when to use this tool versus alternatives. There is no mention of prerequisites, context, or exclusions, leaving the agent without a decision framework.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Close the current Browserbase session
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, and the description only says 'Close', which suggests a destructive action. It does not disclose consequences like invalidating the session or being irreversible, which is needed for safe usage.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single, concise sentence with no wasted words. It efficiently communicates the core purpose.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a simple tool with no parameters or output schema, the description covers the essential purpose. However, it could mention that the session becomes unusable afterward.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
There are no parameters (0 params, 100% schema coverage), so baseline is 4. The description adds no parameter info, but none is needed.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool's purpose: closing the current Browserbase session. It uses a specific verb ('Close') and resource ('session'), and distinguishes from sibling tools like 'start' which begins a session.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description implies when to use the tool (when ending a session), but does not explicitly state when not to use it or mention alternatives. Given the siblings, it's clear, but lacks explicit guidance.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Extract data from the page
| Name | Required | Description | Default |
|---|---|---|---|
| instruction | No |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations and a terse description, the tool's behavioral traits are opaque. It does not disclose whether extraction is read-only, modifies state, requires authentication, or has side effects. The description only states its action, missing critical context for safe invocation.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
While the description is extremely short, it fails to provide necessary details. Conciseness should not sacrifice completeness; here, the sentence is too terse to be useful, especially given the absence of parameter descriptions and annotations.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the low schema coverage, no output schema, no annotations, and a vague description, the tool definition is incomplete. An agent cannot reliably determine how to invoke this tool or what to expect as output, making it nearly unusable without prior context.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The single parameter 'instruction' lacks any description in the schema (0% coverage) and is not explained in the tool description. The agent has no idea what format or content the instruction expects (e.g., CSS selector, natural language). This is a severe gap.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description states 'Extract data from the page,' which clearly indicates a read operation to retrieve data. However, it does not specify what kind of data (e.g., text, attributes, structure), making it vague compared to sibling tools like 'observe' which may also involve data extraction. The purpose is understood but lacks precision.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
No guidance is provided on when to use 'extract' versus alternatives like 'observe' or 'act'. The description does not indicate prerequisites (e.g., page must be loaded), nor does it mention when extraction is appropriate. This leaves the agent to infer from context.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Observe actionable elements on the page
| Name | Required | Description | Default |
|---|---|---|---|
| instruction | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries the full burden. It merely states 'observe actionable elements' without disclosing behavioral traits such as whether it returns data, triggers side effects, or requires permissions. This is insufficient.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is extremely short but lacks substance, making it under-specified rather than concise. It fails to convey necessary information efficiently.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
With one parameter, no schema description, no annotations, and no output schema, the description is entirely inadequate. It does not explain what the tool returns or how to use the instruction parameter effectively.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The only parameter 'instruction' has 0% schema documentation coverage and the description adds no information about its meaning, format, or usage. The parameter is left completely unexplained.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
Description 'Observe actionable elements on the page' clearly states the verb (observe) and the resource (actionable elements). However, it does not differentiate from sibling tools like 'extract' which also reads content, so purpose is clear but lacks distinction.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
No guidance on when to use this tool versus siblings such as 'act' or 'extract'. There are no explicit usage contexts or exclusions provided.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Create or reuse a Browserbase session
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Without annotations, the description carries the full burden. It mentions 'create or reuse' but does not explain the behavior in detail (e.g., what happens on multiple calls, side effects, or lifecycle). The description is too shallow for a tool that manages a session.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is extremely concise—just five words. It front-loads the core purpose. However, it could be slightly more descriptive without added length, e.g., mentioning that it initializes a new session or resumes an existing one.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a tool with no parameters and no output schema, the description is minimally viable. It explains the basic function but lacks context on session lifecycle, error handling, or implications of 'reuse.' More detail would improve completeness.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
There are no parameters, so the description does not need to add parameter semantics. Per the rules, zero parameters warrant a baseline of 4. The description is sufficient for this context.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states that the tool creates or reuses a Browserbase session. It distinguishes itself from sibling tools (act, end, etc.) which are actions within a session. However, it does not explain what 'Browserbase' is or what a session entails, leaving some ambiguity.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
No guidance is provided on when to use this tool versus alternatives. It is implied that you use it before other session-dependent actions, but there is no explicit when/when-not/alternatives statement.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections. Dates show when Glama detected each change.
v3.0.0actendextractnavigateobservestartEach tool targets a distinct operation: session management (start, end), navigation (navigate), interaction (act), data extraction (extract), and element observation (observe). No overlap in purposes.
All tool names are single verbs in lowercase, following a consistent imperative style. No mixing of conventions or naming patterns.
6 tools is well-scoped for a browser automation session server, covering the full lifecycle from creation to teardown, with core actions and data extraction.
Covers the essential session lifecycle and common operations. Missing specific actions like click or type, but 'act' is generic enough. A screenshot tool would be a minor addition.
Unclaimed servers have limited discoverability.
If you are the server author, to access and configure the admin panel.
Provides cloud browser automation capabilities using Stagehand and Browserbase, enabling LLMs to i…
Stealth web automation for AI agents. Login, signup, navigate, screenshot.
Stealth web automation for AI agents. Login, signup, navigate, screenshot.
Headless browser primitives for AI agents when sites need real JS rendering.
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/ai-integr8tor/mcp-server-browserbase'
If you have feedback or need assistance with the MCP directory API, please join our Discord server