Give every AI agent its own identity
An agent on a person’s key inherits all of their access and hides behind their name in every log, so Statio makes each agent a principal of its own, with its own grants, token and off switch.
The borrowed key
Most agents run on a person’s key
Someone pastes their own token into an agent’s config to get it working. From then on the agent can do whatever that person can, every call looks like the person made it, and the only way to stop the agent is to revoke the person.That breaks the three things identity is for
You cannot give the agent less access than its creator. You cannot tell from a log whether a person or an agent acted. And you cannot switch off one misbehaving agent without locking out a colleague.
Borrowed key or own identity
Who is calling, and what stops it
dana@acmeperson · token: dana@acme
log: dana@acme
triage-botagent · token: triage-bot
suspended
deploy-botagent · token: deploy-bot
log: deploy-bot
triage-bot misbehaves again. Suspend triage-bot and it gets no new tokens. Dana and deploy-bot keep working.
| On a person’s key | Its own identity in Statio | |
|---|---|---|
| Access | Everything the person can do | Only what the agent is granted, per server and per tool |
| Activity log | Names the person | Names the agent |
| Stopping it | Revoke the person’s token | Suspend or revoke the agent; the person keeps working |
| Token lifetime | Often a long-lived personal key | Agent tokens last five minutes |
| Upstream API keys | Usually in the agent’s config | Stay in Statio, attached inside the gateway |
How it works in Statio
An agent is a principal
Create an agent in your organization and it sits next to your members: it can be granted servers, granted single tools, and denied tools, exactly like a person.Its own short-lived token
An agent authenticates with its own token, which lasts five minutes. Suspend the agent and it gets no new ones. Nobody else is affected.A status you can see
Each agent shows whether it is active, suspended, revoked or expired, so a list of agents doubles as a list of what can act in your organization.Named in every record
The Activity log records each call’s caller as a user, an agent or unattributed. When something goes wrong you can say which agent did it, and what else it called.
What an agent may call is set with per-tool permissions, and everything it does is in the audit trail.
Questions
You can, and for one agent and one API it works. It multiplies quickly: every agent needs an account in every API, each with its own key to store and rotate, and no single place shows what an agent can reach. A gateway lets an agent hold one identity and keeps the upstream keys in one vault.
Yes. Grant both the server and deny different tools to each. Your release engineers can keep merge_pull_request while the triage agent on the same server cannot call it.
Plans are flat and never per seat. See the pricing page for each plan’s member limits and what it includes.
Lock down your first tool in minutes.
Three MCP servers, one of them hosted, and 10,000 calls a month, free and without a card. If you outgrow it the next plan is $19.
Two steps: create your account, then install the desktop app on each machine that runs an AI client.
- Keys encrypted at rest, never handed to an agent
- Cancel anytime, nothing is deleted
- How Statio handles credentials