Scout
Give your agent a full browser control plane for prototypes, UI checks, and scripted flows without maintaining separate Playwright scripts.
Overview
Scout is an MCP server for the Build phase that provides 84 CDP-based browser automation tools with DOM diffing and vision hybrid support.
What is this MCP server?
- 84 MCP tools for navigation, interaction, extraction, and observation in one Scout server
- Pure Chrome DevTools Protocol (CDP) automation with DOM diffing for stable, token-efficient page state
- Vision-hybrid mode when selectors or layout shifts break classic DOM-only strategies
- Shipped as OCI image ghcr.io/felixgeelhaar/scout:1.9.0 with stdio via mcp serve
- Server version 1.9.0
- Transport: stdio via OCI packageArguments mcp serve
Community signal: 3 GitHub stars.
What problem does it solve?
Indie builders struggle to keep one-off browser scripts in sync when their agent needs reliable click-type-scrape loops across staging and production UIs.
Who is it for?
Builders shipping web SaaS or marketing sites who want agent-driven E2E exploration without maintaining a separate Playwright repo.
Skip if: Mobile-native-only products or teams that forbid browser automation in CI for compliance reasons.
What do I get? / Deliverables
After Scout is registered, your agent can run structured browser sessions with diffed DOM state so you validate flows faster and file sharper bug reports.
- Agent-controlled browser sessions across 84 tools
- Diff-aware DOM snapshots for stable automation
- Containerized MCP endpoint at Scout v1.9.0
Recommended MCP Servers
Journey fit
Browser automation first appears when you integrate real UI behavior during build, then carries into validate prototypes and ship testing. Integrations subphase covers wiring CDP-backed browser tools into the agent stack alongside your app codebase.
How it compares
Full browser MCP with 84 tools, not a lightweight single-tool screenshot skill.
Common Questions / FAQ
Who is Scout for?
Solo builders and small teams using AI agents who need dependable Chrome automation for web apps and landing pages.
When should I use Scout?
Use it while integrating UI flows in build, during validate prototypes, or in ship-phase testing when you need real-browser verification.
How do I add Scout to my agent?
Pull ghcr.io/felixgeelhaar/scout:1.9.0, run the container with mcp serve for stdio transport, and add that command to your MCP server config in Claude Code or Cursor.