Blog

FusionAuth 1.69 adds an AI agent entity type and lifecycle webhooks. Whether that answers the question your auditors ask depends on whose events they are.

MCP standardized how an agent asks for more permission. It has no vocabulary yet for asking whether the human behind the token is still there.

Instagram now limits the reach of AI profiles that skip its AI-generated label. Non-human disclosure just became an enforced account attribute with penalties.

Okta's Agent SSO went GA on the Cross App Access standard. The IdP brokers the connection, but your app still has to validate the grant and issue the token.

How WorkOS engineers make agents babysitting agents safe: a headless surface to supervise through, scoped jobs, visible costs, and a human on the merge button.

Relay proxies an agent's third-party API calls and injects the credential at the boundary, so prompt injection has no token to steal and nowhere to send it.

Which providers rotate refresh tokens, which return expires_in, which give you a grace period, and which revoke on reuse. One row per provider, verified against provider documentation in August 2026.

Concurrent refreshes don't just fail. They can disconnect the user entirely. Here are four layers of defense, cheapest first, and why the Redis lock everyone reaches for isn't the one keeping you safe.

Field notes from running the WorkOS internal-tools showcase twice: five talks, live demos only, one Q&A, a sequenced lineup, and a contingency for every demo.

A walk through the Pipes credential vault: envelope encryption, key context, what a compromised key reaches, and how key rotation happens without re-consent.

Using an Android client and a Node API, including the JWKS failure mode that quietly returns the wrong status code

The path for adding a provider yourself, whether it speaks OAuth, odd OAuth, or only API keys.

Internal AI moved from dashboards into Slack threads. That changes addressing, mandate, presence, and audience — questions that used to apply only to people.

Internal AI teams walk three stages: one hand-built agent, then several, then a governance problem. Stage three needs a substrate — you can't design it first.

The permission prompt breaks down once an agent works continuously. What replaces it is curation: a surplus of candidates, a queue, and a human who picks.