@caelo-cms/mcp-server

MCP server exposing your Caelo CMS install's chat-runner as a single caelo_chat tool. Lets Claude Code (or any MCP-aware client) drive your install — read pages, propose edits, queue Owner-approval proposals, summarise plugin data — without opening the browser.

This package is a thin shim. Every caelo_chat invocation becomes one HTTPS POST against your admin install's /api/mcp/chat endpoint, which dispatches into the same chat-runner that powers the live-edit overlay.

Install

In your Caelo install:

1. Visit /security/mcp as an Owner.
2. Click New token, give it a name (claude-code, laptop, ci, etc.), optionally set an AI-spend cap (microcents), and copy the bearer that's shown ONCE.
3. The page renders the exact claude mcp add snippet — copy + run it.

Manual setup if your client isn't Claude Code:

# stdio server invoked by your MCP-aware client
bunx @caelo-cms/mcp-server

with these env vars set:

| Variable | Required | Notes | |---|---|---| | CAELO_ADMIN_URL | yes | https://admin.example.com — point at your install. | | CAELO_MCP_TOKEN | yes | Bearer minted at /security/mcp. |

The caelo_chat tool

| Field | Type | Notes | |---|---|---| | message | string, required | What you want to say to the Caelo agent. | | chatSessionId | uuid, optional | Continue an existing chat session. | | pageId | uuid, optional | Bind a NEW chat to one page so the agent's page-context block populates. |

Output: assistant reply text + a JSON block with chatSessionId (for the next call), requestId (click through to /security/audit/<requestId> for the full audit trail), the structured tool-call summaries the agent dispatched, the per-turn cost in microcents, and a pendingProposals count so the agent can surface "you have N things waiting for Owner approval".

Why a single tool?

Browse / publish / propose actions happen through the chat — "which pages exist?" → the Caelo agent calls pages.list internally → answers in text. Same auth surface, same RLS scoping, same audit trail, same propose-execute Owner gate as the browser. The remote agent talks to a human-equivalent agent, not to a programmatic API.

License

MPL-2.0.