Scopebond ProductIntegrationsPricingTrustInstalllive decisions, no side effects

Try it

Watch Scopebond decide on an AI agent's action, live.

Pick a team below and choose what its agent tries to do. The open-source Scopebond gateway checks the action against that team's rules, allows or blocks it, and signs the record. These are real decisions with no side effects: nothing here connects to or changes any outside system.

Pick what the agent tries to do:

Prevent

What you just saw is the gateway's live allow-or-block decision. Connect it where your agent runs and refused means it never happens.

Prove

Each decision is signed, so its record can be checked by anyone without trusting us. It proves what was requested and decided, not that a business result happened.

See it as a team

The hosted workspace shows every agent's actions explained in plain language, with a monthly report you can forward. Open a sample workspace — fictional activity, no account.

Give your own agents rules, checked before every action

Free for one agent. Open source, runs in your environment, no card, no sales call.

For developers

Each team is a policy evaluated live by the same gateway. Example:

curl -sX POST https://try.scopebond.com/v1/case/coding/evaluate -H 'content-type: application/json' -d '{"intent":{"action_type":"git.push","params":{"ref":"main","force":"true"}}}'

Inspect: /v1/attester · /.well-known/jwks.json · /healthz
Run it yourself: npx @scopebond/hook init for a coding agent, or npx @scopebond/gateway ./policy.json for any agent over HTTP · source and docs · npm package

Why the records here say insecure_development. This page evaluates intents nobody signed — it is a public demo with no enrolled agent key — so the receipt records that no agent signature was verified, which is the truth about this demo rather than about the product. Your own install enrols a machine key on the first run and records authenticated.