Connect an AI agent

Your workspace can be run by an AI agent — a coding assistant like Claude Code or Codex — so you can ask, in plain language, to add a customer, draft a page, pull a report, or update records. One button in your account generates a key and hands you a paste-ready config for your agent. The agent works with your permissions and nothing more, and every action it takes is recorded. An agent you connect never uses a paid seat.

Connect an agent in one click

1. Click your avatar (top-right) and choose Profile Settings.
2. Open the API & AI Agent tab.
3. Click Connect an AI agent and re-enter your password when prompted.
4. Copy one of the three configs shown — Claude Code, Codex, or JSON. The key is shown once and is valid for one year, so copy it now.
5. Paste it into your agent. A good first message is "call platform_info and tell me what this workspace can do."

That is the whole setup. The dialog builds the config against your workspace's own address, so it is correct whether you sign in at a .everjust.app subdomain or your own custom domain.

Generate a key by hand

If you would rather wire it up yourself, or you are setting up a dedicated agent user:

1. Open Profile Settings → API & AI Agent (the same tab).
2. Choose New API Key, name it (for example "Claude Code"), and copy it — it is shown only once.
3. Paste it into one of the configs below.

For shared or automated use, ask an administrator to create a dedicated agent user with only the access it needs, and mint the key under that user.

The configs

Give your agent your workspace address and the key. Replace <workspace> with your own address and <API_KEY> with the key you copied.

claude mcp add --transport http --scope user everjust \
  https://<workspace>.everjust.app/mcp \
  --header "Authorization: Bearer <API_KEY>"

# ~/.codex/config.toml
[mcp_servers.everjust]
url = "https://<workspace>.everjust.app/mcp"
bearer_token_env_var = "EVERJUST_API_KEY"
# shell: export EVERJUST_API_KEY=<API_KEY>

{ "mcpServers": { "everjust": {
    "type": "http",
    "url": "https://<workspace>.everjust.app/mcp",
    "headers": { "Authorization": "Bearer <API_KEY>" } } } }

What the agent can and can't do

• It works with your permissions. The agent can do what your account can do in the workspace, and nothing your account can't.
• It can't take over the workspace. Granting the Administrator role is refused, from either direction. So are edits to access rules, page templates, scheduled actions, and workspace configuration.
• It can't read your secrets. Stored passwords and secret keys come back masked.
• Deletes are shown first, then confirmed. The agent gets a preview of what would be removed and has to confirm before anything is deleted. Bulk edits over 100 records also need a confirmation.
• Every action is logged. Each call is written to an audit trail an administrator can review; entries are kept 90 days and can't be edited.

One agent, one workspace

Each connection is bound to one workspace. To connect a second workspace, add another named connection (for example everjust-sales) with that workspace's own key.

If it doesn't connect

• The key goes into the command itself. No separate sign-in screen appears; if your agent waits for a browser login, the header or the key is wrong.
• Confirm it worked by running claude mcp list — the server shows as connected and lists its tools — or by asking the agent to run platform_info.

Revoke a key

Open Profile Settings → API & AI Agent, find the key in the list, and delete it. The agent loses access at once. Keys also expire on their own one year after you create them.

Related

• Security & your data — how your workspace protects your data.
• Pricing & billing — agents never count as seats.
• Developers: see AI agents & MCP for the full tool list and the API underneath.