Use cases

Every agent bounded. Every decision provable.

AI agents need read access, write access, and sometimes infrastructure access. Without a policy gate, the only thing separating allowed from unallowed is hope. These are the agent patterns organisations govern first.

DevOps Agent

Deploy to staging freely. Prove production was blocked.

The risk without CLAIIM

A deployment agent with production access is one misconfiguration away from an outage. Without runtime control, you cannot prove what actions were allowed before they happened.

What CLAIIM allows

Deploy to staging, roll back staging, check service health

What CLAIIM blocks

Deploy to production, roll back production, destroy infrastructure

Proof in Chron

Every staging deploy is ALLOW in Chron with agent, anchor, and skill version. Every production attempt is DENY with the policy that fired. Your security team can review the complete record.

View technical example →

Code Review Agent

Read and comment. Never merge.

The risk without CLAIIM

An AI agent reviewing pull requests needs read access to code. If it can also approve or merge, a single error in judgment ships broken or malicious code without human sign-off.

What CLAIIM allows

Read diffs, post inline comments, read pull request metadata

What CLAIIM blocks

Merge pull requests, approve code, push directly, delete branches

Proof in Chron

Every comment is ALLOW in Chron with the engineering lead as accountability anchor. Every merge attempt is DENY. Auditors see the agent could comment but could never ship.

View technical example →

Support Agent

Draft responses. Never send. Never refund.

The risk without CLAIIM

Support agents need to read tickets and prepare responses. If they can also send responses or issue refunds, a bad draft or a classification error becomes a real customer action.

What CLAIIM allows

Read tickets, draft responses for human review, tag tickets

What CLAIIM blocks

Send responses directly, issue refunds, process billing changes, close accounts

Proof in Chron

Every draft is ALLOW. Every billing action is DENY. The support lead is the accountability anchor in every Chron entry -- visible to regulators and auditors.

View technical example →

Finance Reporting Agent

Read reports. Never approve payments.

The risk without CLAIIM

Finance agents need reporting access to produce summaries and forecasts. If they can also approve payments or move funds, a prompt injection or model error becomes a financial incident.

What CLAIIM allows

Read reports, read budget data, generate summaries

What CLAIIM blocks

Approve payments, initiate transfers, modify financial records

Proof in Chron

Finance AIORG is a separate organizational boundary. Engineering agents cannot reach financial data at all. Every denied payment attempt is in Chron with the CFO as accountability anchor.

View technical example →

Privileged Infrastructure Agent

Infrastructure changes only under two-person approval.

The risk without CLAIIM

Infrastructure automation agents can destroy environments, drop databases, or modify clusters. Without a pre-approval gate, the blast radius of an agent error or prompt injection is unlimited.

What CLAIIM allows

Apply pre-approved change orders, read infrastructure state

What CLAIIM blocks

Destroy resources, drop or modify schemas, act without prior human approval

Proof in Chron

Every action inside an approved session is tagged in Chron with the session ID and the approving admin. Direct evidence for EU AI Act Article 14 human oversight. No agent runs without prior approval on record.

View technical example →

See the gate in your environment.

Install guide Book walkthrough