Orkestia for Codex

Orkestia for Codex is the public MCP server inside OpenAI Codex. Codex asks what your organization can run, fills the schema, and starts the workflow. Orkestia records who ran it. The announcement is Orkestia now runs on Codex. This post is the setup.

Orkestia is the backbone that connects software, AI, and the real world. Three ways to call it: the console, the API, and any client that speaks MCP. Codex is the third way, sitting next to the editor you already use to write the change.

Add the server

The URL you paste is:

https://mcp.orkestia.dev/mcp

In a client that reads a JSON MCP config, the entry looks like this:

{
  "mcpServers": {
    "orkestia": {
      "url": "https://mcp.orkestia.dev/mcp"
    }
  }
}

Codex's own settings screen takes the same URL as a streamable HTTP server. Sign in when it prompts you. From that point the server knows which organization you belong to. You never pass an organization id by hand.

Confirm who you are

The first call is always:

whoami()

It returns the principal the server resolved from your token. Two reasons this is worth doing before anything else. It tells you the connection is real rather than merely configured. And it tells you which identity your runs are about to be recorded under.

Find a type, then run it

You do not need the workflow name in advance.

list_workflow_namespaces()
list_workflow_types(prefix="aws.s3.")
get_workflow_schema("aws.s3.create_bucket")
start_workflow(workflow_type="aws.s3.create_bucket", initial_data={...})
watch_workflow(workflow_id)

get_workflow_schema returns the input fields, which of them are required, and a read_only flag. A read-only type is the right first run. Anything else deserves a second look at the schema.

If the schema has has_prerequisites: true, call get_workflow_prerequisites before start_workflow. Connections are the usual case: a cloud, GitHub, Omie, Neon. The guide comes back with Orkestia's side already filled in. The secret stays on the connection. Codex does not see it.

Keep type and id straight. A workflow type is a capability. A workflow id is one execution that belongs to you. Status, history, retry, and terminate all take the id.

What Codex is good for on this catalog

Code that just landed in a repo is the start of a run, not the end of one. From Codex you can open a pull, comment, provision a runner, publish a site with SPAD, or post a title in Omie, using the same identity you would use in the console. The history is how you show someone what happened.

The generic walkthrough, with the status and retry table, is Connect your AI tool, then run your first workflow. Claude and Cursor use the same URL: Orkestia for Claude, Orkestia for Cursor.

Account

Create an account and subscribe. One published plan, same rates for everyone, amounts on pricing.

Docs: Connect an AI assistant. Stuck: hello@orkestia.dev.