# Send BrowserMCP Skill to your agent Hand the extracted package to your coding agent with a concrete install brief instead of figuring it out manually. ## Fast path - Download the package from Yavira. - Extract it into a folder your agent can access. - Paste one of the prompts below and point your agent at the extracted folder. ## Suggested prompts ### New install ```text I downloaded a skill package from Yavira. Read SKILL.md from the extracted folder and install it by following the included instructions. Tell me what you changed and call out any manual steps you could not complete. ``` ### Upgrade existing ```text I downloaded an updated skill package from Yavira. Read SKILL.md from the extracted folder, compare it with my current installation, and upgrade it while preserving any custom configuration unless the package docs explicitly say otherwise. Summarize what changed and any follow-up checks I should run. ``` ## Machine-readable fields ```json { "schemaVersion": "1.0", "item": { "slug": "browsermcp", "name": "BrowserMCP Skill", "source": "tencent", "type": "skill", "category": "开发工具", "sourceUrl": "https://clawhub.ai/frankausberlin/browsermcp", "canonicalUrl": "https://clawhub.ai/frankausberlin/browsermcp", "targetPlatform": "OpenClaw" }, "install": { "downloadUrl": "/downloads/browsermcp", "sourceDownloadUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=browsermcp", "sourcePlatform": "tencent", "targetPlatform": "OpenClaw", "packageFormat": "ZIP package", "primaryDoc": "SKILL.md", "includedAssets": [ "SKILL.md", "templates/config-template.json", "references/setup.md", "references/tools.md", "references/workflows.md", "references/best-practices.md" ], "downloadMode": "redirect", "sourceHealth": { "source": "tencent", "slug": "browsermcp", "status": "healthy", "reason": "direct_download_ok", "recommendedAction": "download", "checkedAt": "2026-05-01T02:41:09.009Z", "expiresAt": "2026-05-08T02:41:09.009Z", "httpStatus": 200, "finalUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=browsermcp", "contentType": "application/zip", "probeMethod": "head", "details": { "probeUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=browsermcp", "contentDisposition": "attachment; filename=\"browsermcp-1.0.0.zip\"", "redirectLocation": null, "bodySnippet": null, "slug": "browsermcp" }, "scope": "item", "summary": "Item download looks usable.", "detail": "Yavira can redirect you to the upstream package for this item.", "primaryActionLabel": "Download for OpenClaw", "primaryActionHref": "/downloads/browsermcp" }, "validation": { "installChecklist": [ "Use the Yavira download entry.", "Review SKILL.md after the package is downloaded.", "Confirm the extracted package contains the expected setup assets." ], "postInstallChecks": [ "Confirm the extracted package includes the expected docs or setup files.", "Validate the skill or prompts are available in your target agent workspace.", "Capture any manual follow-up steps the agent could not complete." ] } }, "links": { "detailUrl": "https://openagent3.xyz/skills/browsermcp", "downloadUrl": "https://openagent3.xyz/downloads/browsermcp", "agentUrl": "https://openagent3.xyz/skills/browsermcp/agent", "manifestUrl": "https://openagent3.xyz/skills/browsermcp/agent.json", "briefUrl": "https://openagent3.xyz/skills/browsermcp/agent.md" } } ``` ## Documentation ### BrowserMCP Skill This skill enables MCP Clients to automate browser interactions through the BrowserMCP protocol. It leverages a local MCP server and a Chrome extension to control the user's actual browser session, allowing for authenticated actions and bypassing common bot detection. ### Key Features ⚡ Fast: Automation happens locally without network latency 🔒 Private: Browser activity stays on the device 👤 Logged In: Uses existing browser profile with active sessions 🥷 Stealth: Avoids basic bot detection and CAPTCHAs via real browser fingerprint ### Prerequisites Before using BrowserMCP automation: MCP Server: BrowserMCP server must be running (via npx) Chrome Extension: Browser MCP Chrome extension must be installed and connected Active Tab: The target browser tab must be connected via the extension [!IMPORTANT] The AI can only control tabs that are actively "Connected" via the Browser MCP extension. If you switch tabs, you must reconnect the new tab. ### Core Workflow The standard browser automation process follows an iterative approach: flowchart TD A[Navigate to URL] --> B[Take Snapshot] B --> C[Identify Elements] C --> D[Interact: Click/Type/etc] D --> E[Wait for Changes] E --> B E --> F[Verify: Screenshot/Logs] ### Standard Pattern StepToolPurposeKey Consideration1navigateOpen the target URLEnsure extension is connected first2snapshotCapture ARIA tree to identify interactive elementsRefresh after any page changes3click / typeInteract with page elementsUse exact ref from snapshot4waitPause for dynamic content to loadEssential after navigation/clicks5screenshotVisually verify resultsUse when uncertain about state6get_console_logsDebug JavaScript errorsCheck when interactions fail ### Essential Tools ToolUse WhenParametersnavigateOpening a new pageurl - full URL including protocolsnapshotUnderstanding page structureNone - returns ARIA accessibility treeclickActivating buttons/linkselement (description), ref (exact ID from snapshot)typeFilling input fieldselement, ref, text, submit (optional)hoverTriggering hover menuselement, refselect_optionChoosing from dropdownselement, ref, values (array)press_keyKeyboard shortcutskey - e.g., "Enter", "Escape", "ArrowDown"waitAllowing page to loadtime - seconds to waitscreenshotVisual verificationNone - returns PNG imageget_console_logsDebugging errorsNone - returns recent console outputgo_back / go_forwardNavigation historyNone ### Common Key Values Navigation: Enter, Escape, Tab Editing: Backspace, Delete Arrows: ArrowUp, ArrowDown, ArrowLeft, ArrowRight Modifiers: Control, Alt, Shift, Meta (combine via modifiers array) Function: F1-F12 Other: Home, End, PageUp, PageDown, Space ### Example 1: Search on Google // Step 1: Navigate to search engine navigate(url="https://google.com") // Step 2: Type search query (use snapshot to find the ref) type(element="Google search box", ref="e12", text="BrowserMCP automation", submit=true) // Alternative: Type then press Enter separately type(element="Search box", ref="e12", text="BrowserMCP automation") press_key(key="Enter") ### Example 2: Fill and Submit a Login Form // Step 1: Navigate to login page navigate(url="https://example.com/login") // Step 2: Get snapshot to identify form fields snapshot() // Step 3: Fill username field type(element="Username or email field", ref="e5", text="user@example.com") // Step 4: Fill password field type(element="Password field", ref="e7", text="password123") // Step 5: Click login button click(element="Sign in button", ref="e9") // Step 6: Wait for redirect wait(time=2) // Step 7: Verify successful login with screenshot screenshot() ### Example 3: Navigate and Extract Information // Navigate to documentation navigate(url="https://docs.browsermcp.io") // Wait for page load wait(time=1) // Capture accessibility tree to understand structure snapshot() // Click on a documentation link click(element="API Reference link", ref="e15") // Wait for content to load wait(time=1) // Take screenshot for verification screenshot() // Check for any JavaScript errors get_console_logs() ### Example 4: Handle Dynamic Content // Navigate to page with dynamic content navigate(url="https://example.com/dashboard") // Wait for initial load wait(time=2) // Take snapshot to see available elements snapshot() // Click element that triggers dynamic content click(element="Load more button", ref="e22") // Wait for new content to appear wait(time=2) // Refresh snapshot to see new elements snapshot() // Interact with newly loaded element click(element="New item", ref="e45") ### 1. Always Use Snapshots for Element Identification Good: // Take snapshot first, then use exact refs snapshot() click(element="Submit button", ref="e12") Bad: // Guessing selectors without snapshot click(element="button.submit") // May not work with dynamic DOM ### 2. Wait After Navigation and Major Actions Dynamic web applications often load content asynchronously. Always wait after: Navigation to a new page Clicking buttons that trigger requests Submitting forms click(element="Load data button", ref="e8") wait(time=2) // Wait for data to load snapshot() // Then get fresh page structure ### 3. Handle Connection Requirements Before any automation: Verify the Browser MCP extension is installed Ensure the target tab is connected (user must click "Connect") If connection errors occur, remind the user to reconnect ### 4. Use Screenshots for Debugging When interactions fail: // Take screenshot to see current page state screenshot() // Check console for JavaScript errors get_console_logs() // Re-snapshot to see updated element refs snapshot() ### 5. Respect Privacy and Security BrowserMCP uses the user's actual browser profile Be cautious with sensitive data User remains logged into their services All actions happen locally on the device ### Reference Navigation FileContentsreferences/setup.mdDetailed installation and configuration for MCP server and Chrome extensionreferences/tools.mdComplete tool reference with parameters and detailed examplesreferences/best-practices.mdAdvanced patterns, error handling, and troubleshooting techniquesreferences/workflows.mdCommon workflow patterns (forms, authentication, scraping, etc.) ### Connection Errors Error: "No connection to browser extension" Solution: User must click the Browser MCP extension icon in Chrome toolbar Click "Connect" button on the target tab Only connected tabs can be automated ### Element Not Found Error: Element reference invalid or element not found Solution: Take a fresh snapshot() - the DOM may have changed Use the new ref values from the updated snapshot Dynamic content may require wait() before snapshot ### Action Blocked or Failed Error: Click/type action didn't work as expected Solution: Take a screenshot() to see current page state Check get_console_logs() for JavaScript errors Verify element is visible and enabled in snapshot Check for browser-level popups or security prompts Ensure the page has finished loading ### CAPTCHA or Bot Detection Note: BrowserMCP helps avoid basic bot detection by using the real browser profile. However: Some sites may still present challenges Rate limiting may apply to rapid interactions User may need to manually solve some CAPTCHAs ### Comparison: BrowserMCP vs Playwright MCP FeatureBrowserMCPPlaywright MCPBrowserUser's existing browserNew browser instanceProfileUses existing profile with cookiesIsolated profileAuthenticationAlready logged inMust log in each sessionBot DetectionLower (real fingerprint)HigherMulti-tabOne tab at a timeMultiple tabs supportedBest ForPersonal automation, testing logged-in flowsTesting, CI/CD, isolated sessions ### Tips for Effective Automation Start with navigate + wait + snapshot - Establish baseline page state Use descriptive element names - Helps with debugging and clarity Take screenshots at checkpoints - Visual verification catches issues early Check console logs after errors - JavaScript errors explain many failures Wait strategically - Too short = flakiness, too long = slowness Refresh snapshots after interactions - DOM changes invalidate old refs Use submit=true for forms - Cleaner than separate press_key("Enter") Combine actions efficiently - Group related operations to minimize round-trips ### Resources BrowserMCP Website: https://browsermcp.io Documentation: https://docs.browsermcp.io Chrome Extension: Search "Browser MCP" in Chrome Web Store GitHub: https://github.com/browsermcp/mcp Based on: Playwright MCP ## Trust - Source: tencent - Verification: Indexed source record - Publisher: frankausberlin - Version: 1.0.0 ## Source health - Status: healthy - Item download looks usable. - Yavira can redirect you to the upstream package for this item. - Health scope: item - Reason: direct_download_ok - Checked at: 2026-05-01T02:41:09.009Z - Expires at: 2026-05-08T02:41:09.009Z - Recommended action: Download for OpenClaw ## Links - [Detail page](https://openagent3.xyz/skills/browsermcp) - [Send to Agent page](https://openagent3.xyz/skills/browsermcp/agent) - [JSON manifest](https://openagent3.xyz/skills/browsermcp/agent.json) - [Markdown brief](https://openagent3.xyz/skills/browsermcp/agent.md) - [Download page](https://openagent3.xyz/downloads/browsermcp)