Headless CMS · Built for agents

The headless CMS for vibe writing.

Author content with your team. Plug in the AI agents you already use to run the CMS on your behalf — scoped, auditable, and native to how your content is shaped.

Works with Claude, Claude Code, Cursor, and anything that speaks MCP or REST.
Why CMS Caddie

Your content, your structure, your collaborators — human and AI.

Most CMSes treat AI agents as an afterthought. We built one where agents are first-class collaborators from day one, working alongside your team on the same structured content.

Author with your team

A clean dashboard for writers, editors, and admins. Define custom content structures — blogs, products, guides, case studies — with typed fields, rich content, and publish scheduling.

Plug in any AI agent

Connect Claude, Cursor, or any MCP-capable tool in two clicks via OAuth. The agent instantly understands your custom content structures and can create, retrieve, update, and delete content on your behalf.

📄

Feed it anything

Drop Word docs, PDFs, notes, or screenshots into your agent. CMS Caddie's schema-aware API guides the agent through converting unstructured input into correctly-shaped, published content.

How it works

Three steps from raw ideas to published content.

You keep creative control. The agent handles the mechanical parts — matching your schema, formatting, slugs, scheduling, and publishing.

1

Model your content once

Define content groups and fields in the dashboard. Add optional example hints per field so any agent knows the exact format you expect (prices, slugs, dates).

2

Connect your agent

Paste one URL into Claude or Cursor. Sign in, tick the scopes you're comfortable granting (read, write, publish, delete), and you're connected.

3

Hand it the work

Share an unstructured document, a rough outline, or a Slack thread. The agent follows your structure, drafts the items, and — if you granted the scope — publishes them.

For your team

A CMS that treats structure as a first principle.

Model it once, serve it anywhere — from your Next.js front-end to a marketing email to the agent drafting next week's post.

Content groups & fields

Build collections like blog, products, or testimonials with typed fields — plain text, rich HTML, numbers, images, video, files, and references between groups.

Publish on a schedule

Every item has one UTC published timestamp. Past = live. Future = scheduled. Far-future = draft. One field, no publish/unpublish dance.

Webhooks on every change

Fire webhooks on create, update, publish, and delete. Plug into rebuild pipelines, cache invalidation, or downstream services without polling.

Multi-project control

One account owns many projects. Invite admins per project, rotate keys per project, keep blast radius tight, and download a full backup zip whenever you want.

Direct-to-S3 uploads

Upload media straight to object storage with presigned POSTs — no server-side proxy, no IAM credentials in the browser, no size surprises.

Human API keys

Traditional long-lived keys with read/write scopes for your applications, static site generators, and CI jobs — still first-class.

For agents

Agents that understand your schema — and can't do more than you let them.

Give an AI agent the same access a teammate has: scoped, auditable, and revocable. No prompt-stuffing credentials. No guesswork about what shape a “number” or a “date” should be.

Native MCP server

Connect Claude Desktop, Claude Code, or Cursor to mcp.cmscaddie.com. Every REST route becomes a typed tool the model can call directly.

🔒

OAuth for non-technical users

Paste one URL. Log in through the browser. Pick a project and scopes. The agent finishes the exchange itself — no API keys to copy-paste.

Fine-grained scopes

Separate scopes for content:read, write, delete, publish, and files. Grant just enough to do the job.

Schema-aware by default

Agents discover your content structures through groups.list and follow per-field format hints you set in the dashboard — no hallucinated fields.

Every action audited

Full audit log per key — action, target, status, request ID, timestamp. Inspect from the dashboard or pull it through the API.

Idempotency & dry runs

Every write accepts an Idempotency-Key and a ?dry_run=true mode. Agents can plan before they act and retry without duplication.

Developer experience

Discoverable by design.

Every feature of the API is described in a single manifest. We project it into OpenAPI, llms.txt, and an MCP tool list from the same source — so new routes propagate to every client automatically.

Pricing

One simple plan. Unlimited everything.

No seat counts, no content caps, no surprise metering. One flat price per project.

Standard
$299 /month
per company · per project
  • Unlimited content
  • Unlimited user accounts
  • Unlimited content groups
  • MCP, OAuth, webhooks, audit log
Get started

Start vibe writing in minutes.

Free to start. Bring your team, your existing docs, and the AI agents you already love.

Get started free