URL: /agent-setup --- title: "Give your agent the keys" sidebarTitle: "Agent setup" description: "One line, pasted into any coding agent. It installs the CLI, wires up MCP, audits your website, and starts fixing what it finds." icon: "bot" --- {/* Default page mode, so the shell keeps the sidebar, breadcrumb and TOC and renders the frontmatter title as the h1. This wrapper only scopes the design-system tokens the sections below use. */}
the whole setup
> Fetch and follow the setup instructions for squirrelscan at https://docs.squirrelscan.com/agent-setup/prompt.md
Works in all six below. And most of the ones that are not.
## Rather do it by hand Fair enough. Six agents we actually test against, each with a real guide. Any other MCP client works too: see [MCP clients](/developers/mcp-clients).
Claude Code anthropic Install the plugin and get skills and MCP in one go, or add MCP on its own. Guide → Codex openai One codex mcp add, or a block in ~/.codex/config.toml. Guide → Cursor cursor One-click install, or two lines in mcp.json if you prefer to see the wiring. Guide → opencode opencode Open source, terminal native. Remote MCP in opencode.json. Guide → Copilot github Editor extension and CLI with agent mode. Add squirrelscan as an HTTP MCP server. Guide → Windsurf cognition Agentic IDE with Cascade. Add squirrelscan as an HTTP MCP server. Guide →
## What it can do once it is in
Audit 278 rules across SEO, performance, security, accessibility, and agent experience.
Read Every issue carries a rule id, a severity, the affected URL, and the fix. Structured tool output, not a PDF.
Track Cloud audits keep a per-website issue tracker your agent can read and comment on.
Spend Nothing, locally. Local audits need no account and cost no credits.
Cloud features (AI page analysis, browser rendering, the shared issue tracker, published reports) need an account: run `squirrel auth login`. Free accounts get 500 credits a month, no card.