Manage tokens and explore endpoints from the API page in my.cubepath.com.
Create a token
Open Organization → API Tokens (or use the Manage API Tokens button on the API page) and click Create Token:
- Token Name: a label to identify the token (e.g.
Production Server). - Permissions: leave Full access on, or turn it off and pick individual scopes.
- Allowed IPs (optional): comma-separated IPs or CIDR ranges that may use the token. Leave empty to allow any IP.
WarningThe token is shown only once, right after creation. Copy it immediately; you won't be able to see it again. If you lose it, delete the token and create a new one.
Scope it down
Full access is convenient and rarely what a token needs. Two settings make a leaked token much less useful to whoever finds it:
- Individual scopes. A monitoring script needs read scopes and nothing else. A token that can only read can't delete a cluster.
- Allowed IPs. Pinning a token to the server that uses it means a copy taken from a config file is useless anywhere else.
Create one token for each script or integration rather than sharing one, so revoking a compromised token affects one thing rather than everything.
Authenticate
Send your token with every request using either header:
curl https://api.cubepath.com/projects/ \
-H "Authorization: Bearer YOUR_API_TOKEN"
You can use X-API-Key: YOUR_API_TOKEN instead of the Authorization header.
NoteSend JSON bodies with
Content-Type: application/json.
Base URL
All endpoints live under:
https://api.cubepath.com
Revoke a token
Go back to Organization → API Tokens and Delete any token to revoke its access immediately.
To rotate a token, create the replacement first, switch the integration over to it, then delete the old one. That way there's no window where the integration can't authenticate.
TipToken use shows up in Activity with Auth Method set to API token (or CubeCLI for requests made with the CubePath CLI). Reviewing that column is how you find tokens still being used by something you'd forgotten about.
Agents and MCP
To let an AI agent operate your infrastructure, the MCP server is usually a better fit than handing over a raw token: it has its own authorization flow, per-organization consent and permission checks on every request. Tokens remain the right answer for scripts, CI and anywhere there's no browser to complete a login.