Skip to content

For agents

An API that explains itself.

An MCP server translates for an API that cannot describe itself. This one can.

Start free
Webhook endpoints in the ZevContent dashboard, each with the events it subscribes to and its delivery status
0
MCP servers to install, host or keep in sync
1
call to learn the entire project
1
project a key can be scoped to, with an expiry it can read

Discovery

Read the whole project in one call.

An MCP server exists to translate for an API that cannot describe itself. This one can, so an agent starts working from the response instead of from a wrapper somebody has to maintain.

  • Every content type, field and response shape
  • Every endpoint and every webhook event
  • Generated from what the API actually serves
Introspection
Custom content types in the ZevContent dashboard, the same model an agent reads through introspection

Writing

The same surface a website reads from.

Define types, create entries, upload media and publish, all through the REST API. There is no second, weaker interface for automation.

  • Create and change content types from a script
  • Create, update and publish entries
  • Every refusal names what would have worked instead
Building with agents
Entries in the ZevContent dashboard, each with its publish status

Boundaries

Keys that know when they end.

Scope an agent to one project with a session key that expires in days rather than years, and reports its own expiry so it can stop cleanly rather than fail halfway through a task.

  • Session keys expire; service keys do not
  • One project per key, never a whole workspace
  • Revoke one without touching anything else
Keys and access
API keys in the ZevContent dashboard, including an expiring session key created for a script

Frequently asked questions

What agents ask before they start.

Is there an MCP server?

No, and there does not need to be. An MCP server exists to describe an API that cannot describe itself; one introspection call here returns the whole contract.

Can an agent write, or only read?

Write. Types, entries, publishing and media all go through the same REST surface a website reads from, with a secret key scoped to one project.

How do I stop an agent going too far?

Scope its key to a single project and give it an expiry it can read back, so it knows when it is about to lose access rather than failing halfway through a task.

Start with a project and a key.

Free while we are in beta. Paid plans when we ship them, and never a surprise bill.