How it works
A checkpoint your agents can't argue with, and a record nobody has to take our word for
Think of Scopebond as the approvals desk an agent has to pass through. It doesn't read the agent's conversation, only what it's about to do, and it answers the same way every time: allowed, refused, limited, or "ask a person."
Write the rules once, in plain terms
How much an agent may spend, to whom, which systems it may touch, which hours, what needs a person's OK. Each rule is either enforced (stopped on the spot) or watched (allowed, flagged, and covered by recourse later).
Put the checkpoint where the agent works
Tell us where your agent runs: a coding tool, GitHub, your own software, a cloud, ServiceNow. We generate the setup. Your engineer does it once; you never touch a terminal.
See every action, explained
"Support agent requested a $420 refund, refused by the $200 limit, not carried out, record verified." Name who's responsible, review rules with a colleague, export evidence when asked.
How strong is the protection? We tell you, on every connection.
Different places allow different guarantees. Rather than promise "one click for everything," every integration carries one of four labels.
Scopebond decides before the action runs. Refused means it never happens.
The agent's software checks with Scopebond and follows the answer. Anything built outside it isn't covered.
The work already happened somewhere we can't reach; we stop it being merged, deployed or sent out, and record it.
We keep a signed record of what a platform reports. We can't prevent these, and we say so.
Why a record, not just a log, matters
A log lives in one vendor's dashboard and says what they say it says. A Scopebond record is signed, so anyone, your auditor, your customer, the other side of a dispute, can check it themselves, even with our servers switched off. It's the difference between "trust us" and "see for yourself."
This is the technical receipt behind the decision on our homepage. Press Tamper with it to change one word and watch the signature fail. The check runs in your browser; nothing leaves the page.
{
"type": "scopebond:receipt",
"decision": "deny",
"reason": "amount 18000 > limit 10000",
"action_ref": "sha256:9f2c0b7ad1e4c6f80b3a2d15c7e9f4a1a71b",
"policy_digest": "sha256:41d0e2a9c7b3f6108d4e5a2c9b0f7e13c88e",
"issued_at": "2026-09-16T14:22:07Z",
"issuer_id": "gw_prod",
"kid": "ed25519:3b9f",
"executed": false,
"external_effect": "not_independently_verified"
}Verified against the published Ed25519 key. Offline: npx @scopebond/gateway verify receipt.json
What Scopebond doesn't do
It doesn't read prompts or conversations, judge code quality, or catch manipulation inside the agent. It decides whether a specific action is within your rules, and it proves what was decided. A record proves what was requested and decided, never that a business result happened.
For your engineers
Open source, one command to scaffold, works over HTTP and MCP, verifies offline. The technical docs, receipt format and policy language are on GitHub.
npx @scopebond/gateway initDeveloper quickstart →