Hosted browser automation for agent clients
Scout's browser runtime is the broadest public surface: one hosted environment for clean browser sessions, CDP-backed automation, canvas and WebGL inspection, commerce flows, and multi-step web workflows.
navigateinspectextractsamplesettleGlass, network light, accessibility trees, and a visible browser frame that feels ready to be driven rather than merely observed.
A stack of clean browser chambers: isolated enough for repeatable automation, alive enough for real page behavior, and transparent enough for an agent to reason through every step.
Browser turns the open web into an inspectable runtime for agents.
The browser connector is Scout's widest lens, but its job is not to impersonate a human browsing session. It creates a controlled chamber where pages can be opened, inspected, driven, and verified without inheriting personal state by accident.
Its identity is the instrumented window: familiar web surfaces on top, protocol machinery underneath, and a steady feedback loop between what the agent sees and what it does next.
clean sessions
CDP events
observe, act, verify
Real browser state belongs to the extension path
The public browser connector is the default automation path. When a workflow needs your real Chrome tabs, cookies, and operator-visible state, move to the Scout extension instead of stretching the hosted connector past its boundary.
Use the hosted browser connector for isolated runs, scraping, testing, and reproducible automation.
Use the extension app path only when the agent genuinely needs your existing browser session.
Do not treat the extension as another MCP server config. It has its own app runtime and provider-key setup.
Navigation, extraction, screenshots, forms, downloads, cookies, and network-aware automation.
Canvas and WebGL surfaces that need pixel diffing and point-based interaction beyond the DOM.
Commerce workflows: balances, challenges, and payment execution inside the browser session.
Agent workflows that need the broadest Scout tool surface in one connector.
Runs clean managed browser sessions backed by CDP-aware runtime ownership.
Can launch, attach, inspect, and coordinate sessions without assuming an existing personal browser context.
Keeps browser automation as one surface inside the wider Scout environment instead of standing in for the entire product.
Use OAuth by default for interactive MCP clients and operator workstations.
Switch to Scout MCP tokens only for CI, scheduled jobs, or other environments that cannot open browser sign-in.
Allowance and budget limits are checked at the connector boundary before longer browser runs continue.
A compact map of what this connector can do.
Browser exposes 80 commands across 13 capability groups. The full reference stays close, while this page keeps the connector's runtime shape easy to understand.
Session
9Manage browser sessions, tab attachment, and multi-agent coordination.
browser-tabsbrowser-detachbrowser-launchbrowser-openbrowser-closebrowser-connectbrowser-disconnectbrowser-remote+1 moreNavigation
6Navigate between pages, manage browser history, and handle iframes.
browser-navigatebrowser-searchbrowser-historybrowser-scenebrowser-framesbrowser-waitContent
12Read and understand page content — snapshots, extraction, JavaScript evaluation.
browser-snapshotbrowser-elementbrowser-extractbrowser-evaluatebrowser-pipebrowser-findbrowser-injectbrowser-file+4 moreInteraction
6Simulate user interactions — clicks, keyboard input, form filling, drag-and-drop.
browser-interactbrowser-attachbrowser-dialogbrowser-highlightcanvas-interactcanvas-sequenceNetwork
7Monitor network traffic, intercept requests, record HAR files, manage certificates.
browser-networkbrowser-routebrowser-unroutebrowser-harbrowser-securitybrowser-notificationsbrowser-websocketStorage
4Manage cookies, localStorage, sessionStorage, and clipboard.
browser-cookiesbrowser-storagebrowser-clipboardbrowser-bookmarksMedia
8Capture screenshots, record screencasts, manage downloads, control media playback.
browser-screenshotbrowser-pdfbrowser-screencastbrowser-mediabrowser-downloadbrowser-transcribebrowser-stillbrowser-imageDebug
6Inspect console output, measure performance, monitor DOM memory stats.
browser-consolebrowser-metricsbrowser-memorybrowser-exceptionsbrowser-heapbrowser-traceEmulation
6Emulate devices, geolocation, vision deficiencies, and network/CPU throttling.
browser-emulatebrowser-resizebrowser-visionbrowser-cpubrowser-throttlebrowser-pinchAutomation
5High-level multi-page workflows and batch operations.
browser-batchbrowser-workflowbrowser-crawlbrowser-taskbrowser-animationStyling
1Inspect and manipulate CSS computed styles.
browser-stylesExtension
5Sideload an extension, open its pages, drive the background worker, manage storage, and reload.
extension-launchextension-openextension-backgroundextension-reloadextension-storagePayment
5Read balances, stage challenges, run pay flows, and move funds through the payment connector.
payment-balancepayment-challengepayment-commercepayment-paypayment-transferAuthenticate with a bearer token.
Every Scout connector accepts a bearer token in the Authorization header. Sign in to your Scout account and create an MCP token from the MCP Tokens page, then paste it into your MCP client config.
SnippetJSONKeep connector boundaries intentional
Dedicated connector pages should make capability boundaries memorable before the agent ever receives tools.
