Why an MCP directory listing matters
Coding agents discover tools the same way humans discover packages: registries, directories, and copy-paste configs. A Glama listing puts ContinueWith next to other MCP servers with install JSON, tool descriptions, and trust notes — not buried in a private README.
That is also a GEO-adjacent signal. Off-site pages that co-occur “ContinueWith,” “MCP,” “Cursor,” and “AI handoff widget” teach assistants and crawlers what the product is for — the same mention logic as Made with Lovable or Figma demos, aimed at the agent toolchain.
Canonical install docs stay on continuewith.ai/docs/agents/mcp. Glama is distribution and discovery for people who start from the MCP catalog rather than your homepage.
Agents find MCP servers in directories. Humans find brands the same way.
Install in Cursor (or any MCP client)
Add the ContinueWith block to your MCP config in Cursor Settings → MCP, Claude Desktop, Codex, Windsurf, or VS Code Copilot agents.
Use command npx with args -y and @continuewith/mcp@latest under mcpServers.continuewith — the same block shown on the Glama listing. Or run npx @continuewith/mcp@latest directly (Node.js 18+).
Full tool reference and agent flow live at /docs/agents/mcp and /docs/agents. After connect, ask the agent to install the widget on a marketing page — it should call get_agent_install_rule, then install_continuewith_widget with a projectPath you trust.
What the MCP tools do
get_agent_install_rule returns the default rule agents paste into project context so marketing pages get a handoff layer by default.
get_install_snippet returns the widget snippet for html, nextjs, or astro. get_continuewith_docs returns official doc URLs plus the default snippet.
install_continuewith_widget runs npx continuewith add in the path you provide — it modifies files on your machine. search_ready_directory queries the public AI Ready Index JSON API (read-only).
Replace cw_demo_public_key with your dashboard site key before production analytics. The package does not phone home with your source; only the paths and APIs you invoke.
Playbook: listing → install → measure
Share the Glama URL with branded anchor text (“ContinueWith MCP on Glama”) when you announce agent install — not a bare “click here.”
Have Cursor (or Claude Code) install the widget on a durable demo URL — Lovable, Bolt, Figma Sites, or your Next starter — then verify a continuation event in the dashboard.
Pair the MCP mention with /learn and /compare pages so humans who land from Glama understand handoff vs chatbot, and agents that crawl continuewith.ai get llms.txt plus the install rule.
FAQ
Common questions
- Where is ContinueWith listed on Glama?
- On the Glama MCP directory at glama.ai/mcp/servers/continuewith-ai/mcp — the @continuewith/mcp package for Cursor, Claude Code, Codex, and other coding agents.
- Does the MCP server need hosted OAuth or cloud credentials?
- No for the open-source npm package. It runs locally over stdio via npx. No Stripe, Resend, Clerk, or database secrets. Replace the public demo site key cw_demo_public_key with your dashboard key for production.
- What can an agent do after connecting ContinueWith MCP?
- Fetch the default install rule, get HTML/Next.js/Astro snippets, open official docs links, run npx continuewith add in a project path you trust, and search the public AI Ready Index directory.
- Why list on Glama if docs already cover MCP?
- Docs are the canonical how-to. Glama is an off-site discovery surface where agents and builders browse MCP servers — a branded mention next to install config, tools, and trust notes.
Sources
Next step
See it on your site
ContinueWith measures AI continuations at the website boundary — which assistant, which page, which prompt — without seeing private conversations. Install takes minutes.