Looking Glass Mcp
Give coding agents a full browser stack—navigation, extraction, and credentials—via 71 MCP tools with optional vault CLI and security profiles.
Overview
Looking Glass MCP is a MCP server for the Build phase that gives agents 71 self-healing browser tools, semantic page extraction, and a vault CLI over stdio.
What is this MCP server?
- 71 browser-oriented MCP tools for agent-driven navigation and interaction
- Self-healing automation and semantic extraction for brittle or changing DOMs
- Vault CLI for encrypted credentials: npx looking-glass-mcp vault store <profile>
- Browser modes via BROWSER_HEADLESS plus security profiles: restricted, local-dev, open, sandbox
- npm package looking-glass-mcp v3.1.1 with stdio transport
- 71 MCP browser tools
- Package version 3.1.1
- Four documented security profiles
What problem does it solve?
Agents break on every layout change and cannot safely reuse logins when you need real browser automation inside the coding workflow.
Who is it for?
Solo builders who want local stdio MCP browser automation with security profiles and encrypted credential profiles for repeated agent tasks.
Skip if: Builders who only need static HTTP fetches, cannot run a local browser, or want a fully managed cloud browser with no local ops.
What do I get? / Deliverables
After npm install and env configuration, your agent can browse, extract structured content, and store credentials via vault without one-off Puppeteer scripts.
- Agent-executable browser sessions via 71 MCP tools
- Structured semantic extractions from live web pages
- Encrypted credential profiles managed through vault CLI
Recommended MCP Servers
Journey fit
Agent-native browsing is core build-phase tooling when automating flows, scraping structured pages, and validating UIs from the same environment you code in. Looking Glass is stdio MCP packaged on npm, extending the agent runtime rather than replacing a human-driven browser extension alone.
How it compares
Local browser automation MCP with 71 tools, not a lightweight URL-fetch skill or a passive research-only MCP.
Common Questions / FAQ
Who is looking-glass-mcp for?
Developers using MCP agents who need durable browser control, semantic extraction, and optional encrypted credential profiles on their machine.
When should I use looking-glass-mcp?
Use it when agents must log into sites, navigate multi-step flows, or extract structured data from live pages during build, validate, or ship checks.
How do I add looking-glass-mcp to my agent?
Register the npm package looking-glass-mcp with stdio transport, set BROWSER_HEADLESS and BROWSER_SECURITY_PROFILE, optionally set VAULT_ENCRYPTION_KEY, and store secrets via npx looking-glass-mcp vault store.