How to govern Claude Code
An agent that reads, edits and runs code in your repositories, with the oversight that autonomy demands.
Anthropic · Reviewed when the vendor ships a capability change or a customer declares the tool
How Claude Code is used
We see two common ways businesses use Claude Code, and we govern each one on its own terms. If your team uses it differently, that is fine: you describe the use in your own words and Certrak works out the governance from what it actually does.
Code generation and review
Consequence to people: LowWriting or reviewing code, with a developer in the loop
Internal informationEvery output reviewedInternal use only
Autonomous coding agent
Consequence to people: MediumCode AI operating in agent mode - taking actions across repositories, running commands, deploying
Internal informationMost outputs reviewedInternal use onlyRuns code or scriptsActs on its own
Governing Claude Code in practice
- Name one person who owns Claude Code in your business. Governance starts with a name.
- Write down what Claude Code is used for. Certrak recognises 2 common uses: Code generation and review and Autonomous coding agent. Yours may differ; write down yours.
- Decide what Claude Code may and may not be given. Internal business information is in scope, so name what stays out.
- A person reviews most outputs and exceptions are flagged. Agree what counts as an exception and who sees it.
- Claude Code can run code or scripts without a person approving. Record which of these you allow, and who can change that.
- Look at this again every month, and sooner on any of these: industry incidents, vendor model updates and autonomy level changes.