USE CASE
Agents are stateless. They don’t detect API drift. When a required field disappears or a type changes, the agent fails at runtime — not at build time. StateAnchor catches the drift at the PR level, before it merges.
THE PROBLEM
Traditional software fails loudly when an API breaks. Type errors, compilation failures, integration tests. Agents don’t have those safety nets. An agent that calls a deprecated endpoint will keep trying, often confidently, until a human notices the output is wrong.
The window between “API changed” and “agent fails in production” can be days. By then, the agent has been running against a broken contract with no indication that anything is wrong. StateAnchor closes that window to zero — the drift is caught before the PR merges.
HOW IT WORKS
One file in your repo root declares every endpoint, model, and auth scheme your agents depend on. It becomes the contract.
On every push, the gate diffs the spec against the prior snapshot. Any change that would break an agent caller gets classified as ERR and blocks the merge.
The gate passes only when the change is safe. No drift reaches production. Your agent never encounters a broken API.
WHAT GETS CAUGHT
These are the ERR-lane change kinds that matter most to agent callers. Each one blocks the PR automatically. No configuration required.
ERRThe endpoint your agent calls no longer exists.
Agent fails at runtime with no recoverable path.
ERRA new required field was added to a request the agent constructs.
Agent sends invalid requests; API rejects them with 400 or 422.
ERRA field type changed -- string to integer, or array to object.
Agent sends or receives the wrong shape. Silent data corruption.
ERRThe auth scheme or required scopes changed.
Agent gets 401 or 403 on every call until manually updated.
Connect your repo in ~=5 minutes. The gate runs automatically on every push.
Free to start. No credit card required.