Let your agent set this up

Everything on this site can be done by the agent you already work in. One command connects it, and then you ask in words.

Connect it

# Claude Code
claude mcp add --transport http axela https://axela.app/mcp

# Codex
codex mcp add axela --url https://axela.app/mcp

The first time your agent uses it, a browser opens and you sign in — the same account as this site. There is no token to copy, paste or keep anywhere: the agent gets its own credential and it expires on its own. If you are already signed in here, it takes one click.

Any other MCP client

The endpoint speaks streamable HTTP and advertises where to sign in, so a client that supports OAuth needs only the address:

{
  "mcpServers": {
    "axela": {
      "type": "http",
      "url": "https://axela.app/mcp"
    }
  }
}

A client that cannot sign in through a browser can send a bearer token instead. The sign-in server is published at /.well-known/oauth-protected-resource, which is how a client discovers it without being told.

Then just say it

These are not commands to memorise. They are what people actually type, and the agent works out which calls to make:

Say this	What happens
"Set up an organisation on Axela for my team."	Registers it, points it at your repository, and tells you the next step. You approve the parts that need a person in the browser.
"Publish the skills in this repository."	Reads the publication state, signs from your machine, and confirms the exact catalog that went live. The signing key never leaves your computer.
"What changed on our machines?"	Reads the signed reports and explains each one — what changed, on which computer, and what the computer already did about it.
"Someone left. Close their access."	Walks the whole order: their machines first, then their membership, then any invitation they never used. Nothing of theirs is left enabled.
"Move Anna's laptop to the platform team."	Disables it where it is, invites it where it belongs, and tells you what the person has to run on the laptop itself.

What it can reach

The same things you can, and only for the organisations your account already belongs to. Nothing here is a second set of permissions to keep in step with the first.

Read — your organisations, their connected computers and what each last reported, who has joined and who was invited, what is published from which repository, and every signed event including the skills that no longer match what was published.
Change — register an organisation, add a repository, invite and withdraw invitations, disable a member, disable a computer, rotate an organisation's token.

What it cannot do

Worth saying plainly, because an agent that could do these would defeat the point of the product:

It cannot sign anything. Publishing is signed by a key that stays on your machine. This connection can read publication state and tell you what to do; it cannot approve bytes.
It cannot read the skills installed on your computers. Those copies stay where they are — nothing here can show them to an agent, only names, versions and fingerprints. What you publish is different and is worth being exact about: Axela fetches that repository from GitHub at the commit you published and reads it before signing, which is where the review on a catalog's page comes from.
It cannot reach another organisation. Every call is checked against your account, the same way this page is.

The other half of the setup happens on each computer: a small local checker that verifies installed skills before a session and files the signed report you read here. That part is on the front page, and your agent can walk you through it too.