Last updated August 12, 2026
Liveloop is an MCP (Model Context Protocol) connector, so you can ask any compatible AI client to create and publish loops on your Liveloop account. Your AI writes the project; Liveloop safely hosts and shares it. Claude is one supported example, not a requirement.
No secret to copy or store — choose your client below:
1. Add Liveloop as a remote Streamable HTTP MCP server.
2. Use this connector URL:
https://www.liveloop.space/api/mcp
3. A Liveloop page opens — sign in if needed, then press Allow. You’re connected.
Then just ask, e.g. “create a 3D spinning-cube loop on my Liveloop and publish it.”
codex mcp add liveloop --url https://www.liveloop.space/api/mcp
codex mcp login liveloopclaude mcp add --scope user --transport http liveloop https://www.liveloop.space/api/mcpkimi mcp add --transport http --auth oauth liveloop https://www.liveloop.space/api/mcp
kimi mcp auth liveloop1. In Liveloop, open Settings → Connectors and click Generate token (it’s shown once — copy it).
2. Add the same connector URL and supply the token as a bearer credential:
Authorization: Bearer llk_…
Use the exact Authorization: Bearer header format. Keep the token secret and revoke it if it is ever exposed.
This is separate from MCP. In Settings → API keys, pick OpenAI, Claude, Gemini, DeepSeek or Llama via Groq, open the provider's key page, paste the key and press Connect. Liveloop verifies it without spending generation credits, encrypts it server-side and never exposes it to Loops.
The connector grants scoped permissions to create and publish loops. It can create hosted HTML, Markdown, React or Story loops, or turn a creator-controlled public project URL into a Link Loop after an explicit responsibility confirmation. Liveloop checks Link destinations server-side and decides whether they may run embedded or must stay external. The connected AI can edit, remix, publish, unpublish, search and list, but it cannot delete loops, alter the platform design, read your private data, see other conversations, or act as anyone else.
We store only a hashed connection token and the loops your assistant creates — never your AI conversation. You can revoke a connection at any time in Settings → Connectors or in your AI client's settings; it takes effect immediately. See the Privacy Policy (section 13) for the full detail.
Questions or issues with the connector? Reach the maintainers through the community templates repository. Liveloop itself is proprietary and source-available; the loop templates are open source (MIT) and welcome contributions there.