Decide what the agent may do.
Claude Code runs under your policy through its hooks: each prompt is checked, and each tool call is allowed, blocked or sent to a person. Codex and Gemini CLI report every step.
Agent policy
Claude Code runs under your policy through its hooks: each prompt is checked, and each tool call is allowed, blocked or sent to a person. Codex and Gemini CLI report every step.
- Agents
- Claude Code, Codex
- Allow
- read, edit, npm test, go test
- Ask first
- git push, gh pr create
- Block
- piped installs, rm -rf /, git push --force
- MCP servers
- github, sandscale; others denied
- Prompts
- checked by DLP on Enterprise
- Claude Code
- Enforced through its hooks
- Codex, Gemini CLI
- Recorded over OpenTelemetry
- Sessions
- Pause or stop any session
- Kill switch
- One switch stops every agent in your organisation
- MCP servers
- Allow and deny lists
How it works
What happens, in order.
01Write the policy
Per project: what agents may do, what needs a person, and what is never allowed.
02Enforce it
Claude Code’s hooks ask Sandscale before each tool call and prompt.
03See it
Every decision is in the session record and the audit log.
Give every agent
a box of its own
Start free with 100 vCPU-hours a month. Upgrade when your agents need more.
A Firecracker microVM for every runNo network by defaultPause and resume with state keptVault keeps secrets out of promptsSSO, roles and audit logBring your own cloudWorks over MCP and the CLI