What is Agent Preflight?
It is a bounded, account-scoped, read-only way to check a declared proposed tool, API, webhook or workflow connection before action. It can identify issues such as missing required fields, incompatible formats, an output without a declared consumer, or an assumed adapter that is not present.
What does read-only mean?
It does not change a workflow, configure a destination, sign policy, queue or execute work, activate a system, or make arbitrary network calls. A result is a redacted compatibility finding and a bounded next action for human review.
Does a compatible result mean we can run the workflow?
No. A compatible result is not permission, execution authority, deployment approval or a safety guarantee. Scope, destination and data-boundary decisions remain separate human responsibilities.
Can we send secrets or production payloads?
No. Do not use a public-safe pilot conversation or compatibility check for credentials, private keys, confidential payloads or production secrets.
Is this the same as webhook delivery?
No. Agent Preflight checks a proposed connection before action. Webhook Boundary is the separate Ingress Shield surface for protecting and observing declared webhook traffic on a reviewed route.
How does MCP fit in?
Approved clients can use a protected MCP interface for the bounded read-only tools. Exact live availability and argument schemas are discovered through approved onboarding and tools/list, not copied into a public page.
Is access open to anyone?
No. The pilot is controlled and reviewed; external pilot access is not yet available. It does not claim public self-service enrolment, general availability or a contractual uptime guarantee.
What happens after a result?
A person reviews the stated next action. Agent Preflight does not execute remediation or change the connected systems.