We run AI agents in production for clients: SEO, content, marketing operations, and internal tooling. The parts that work anywhere are open here.
- Skills: https://scalably.io/skills/ (22 production-derived Agent Skills, source in agent-skills)
- MCP servers: eight servers we run in production, published under the
io.scalablyregistry namespace: gsc-mcp, ga4-mcp, google-ads-mcp, hunter-mcp, klaviyo-mcp, figma-mcp, dataforseo-mcp, shopify-mcp. Templates: Python, TypeScript. - Hooks: claude-code-ledger-hooks today, a full hooks plugin next
Every gallery page has a machine-readable form for agents: append .md or .json to the page URL, or request it with Accept: text/markdown.
claude plugin marketplace add scalably-io/agent-skills
claude mcp add gsc -e GOOGLE_APPLICATION_CREDENTIALS=/path/to/service-account.json -- uvx scalably-gsc-mcp
claude mcp add hunter -e HUNTER_API_KEY=... -- npx -y @scalably-io/hunter-mcpPython servers install with uvx scalably-<server>, TypeScript servers with npx -y @scalably-io/<server>; each README has the exact line for Claude Code, Codex and Claude Desktop.
Each open piece is derived from a server or skill we run in production, stripped of anything platform-specific, tested in a clean install, and versioned. Users bring their own API keys. Nothing here phones home.
Questions: hello@scalably.io. Security reports: see SECURITY.md in any repository.