Most AI automation platforms connect to your tools via APIs. That works well for structured data — reading a Jira ticket, pulling a Salesforce contact — but it misses everything that happens in the browser. The email you’re composing in Gmail. The Slack thread you’re reading. The form you’re filling out in ServiceNow.
JieGou’s browser extension bridges that gap. Using the Model Context Protocol (MCP), it gives AI direct access to 60+ browser automation tools — from clicking buttons to reading page content to capturing network traffic.
What is MCP?
The Model Context Protocol is an open standard for connecting AI models to external tools and data sources. Instead of building one-off API integrations for every service, MCP provides a universal interface that any AI model can use to discover and invoke tools.
JieGou’s browser extension implements an MCP client that connects to the JieGou server via WebSocket. When a recipe or workflow needs to interact with a browser-based tool, it sends an MCP tool call through this connection, and the extension executes it in the user’s browser.
What can the extension do?
The 60+ tools fall into several categories:
Page interaction — Click elements, fill form fields, select dropdowns, check boxes, scroll, and navigate between pages. The AI can operate your browser the way you would.
Content reading — Read the visible text on a page, extract specific elements by CSS selector, get the page title and URL, and capture the full DOM structure. This gives AI context about what you’re looking at.
Screenshots — Capture the visible viewport or a specific element. Useful for visual verification of UI state or for feeding into multimodal AI models.
Network capture — Monitor and capture network requests and responses. This helps AI understand what data is flowing between the browser and backend services.
Platform-specific tools — Specialized handlers for Gmail (compose, read, search emails), Slack (read messages, post to channels), Jira (create issues, update tickets), Salesforce (read records, update fields), ServiceNow (manage incidents), and HubSpot (access contacts and campaigns).
The dedicated automation window
When running complex automations, you don’t want the AI clicking around in your active browser tab. JieGou opens a dedicated automation window — an isolated browser context where MCP tools execute without interfering with your work.
This window runs in the background, so automations continue even while you’re using the browser for other tasks. Results are streamed back to the console in real time.
How it connects to recipes and workflows
Browser tools appear alongside all other tools in the recipe editor and workflow builder. When you create a recipe that needs to read data from a web page or fill out a form, you select the relevant browser tool from the tool list.
During execution, JieGou routes the tool call to the browser extension, waits for the result, and feeds it back into the recipe or workflow step. The AI treats browser tools exactly like API tools — there’s no special syntax or configuration.
Platform-specific intelligence
The extension doesn’t just interact with web pages generically. For supported platforms, it has specialized inject scripts that understand the platform’s DOM structure and can extract meaningful data.
For example, the Gmail handler can identify email threads, extract sender and recipient information, read message bodies, and compose new emails — all by understanding Gmail’s specific HTML structure. The Jira handler knows how to read issue fields, comments, and sprint data from Jira’s interface.
This platform-specific intelligence means the AI can work with your tools at a higher level of abstraction than raw DOM manipulation.
Security and permissions
The browser extension requires explicit user authorization and operates within multiple layers of security.
Authentication and authorization. The extension connects to JieGou’s MCP server using JWT-based authentication. Each WebSocket connection is verified against the user’s session, and the extension cannot execute tools without a valid, active token. Tokens are short-lived and scoped to the user’s account permissions.
API key validation. Before any tool call is dispatched, the system validates that the user’s API keys are active and properly configured. Invalid or revoked keys are rejected before reaching the browser, preventing unauthorized automation attempts.
Rate limiting. Per-user and per-tool rate limits prevent abuse and accidental runaway automation. If a workflow triggers an unusually high number of browser actions, rate limiting throttles execution rather than overwhelming the target application.
Audit logging. Every tool call is logged with the user, tool name, timestamp, target URL, and result status. These audit entries are queryable and exportable, giving admins full visibility into what browser automation has been performed and by whom.
Tool ACLs. Administrators can configure allow and deny lists that control which browser tools are available to which roles. A support team might have access to ticket management tools but not to tools that modify CRM records. ACLs are enforced at the server level before tool calls reach the extension.
BYOK compatibility. Bring Your Own Keys applies to browser automation just as it does to LLM calls — the extension operates in your browser, and data flows between your browser and the web applications you’re using. JieGou orchestrates the automation but doesn’t see the page content.
Getting started
The browser extension is available for Chrome and Firefox. Install it from the extension store, sign in with your JieGou account, and browser tools become available in your recipes and workflows immediately. Learn more about browser automation or see all integrations.