Recipes

Guard deployments

For an agent that ships code, in 15 frameworks.

Last reviewed 2026-10-02

Staging deploys run on their own; production waits for a person.

The rule

{
  "name": "deploy-guard",
  "effect": "require_approval",
  "priority": 10,
  "match_path": "/tools/deploy_service",
  "conditions": [
    {
      "field": "environment",
      "op": "eq",
      "value": "production"
    }
  ]
}

Measured on production (2026-10-02, rule tester): the risky call below matches this rule and the safe one does not.

Risky	`{"service": "billing", "environment": "production"}`
Safe	`{"service": "billing", "environment": "staging"}`

Pick your framework

Framework	Tool name CAIN sees	Plugs into
Python (any function)	`deploy_service`	the function that acts
LangChain	`deploy_service`	your agent's tool list
LangGraph	`deploy_service`	the tools your graph's nodes call
OpenAI Agents SDK	`deploy_service`	function tools
CrewAI	`deploy_service`	CrewAI's before-tool-call hooks
LlamaIndex	`deploy_service`	FunctionTool
Pydantic AI	`deploy_service`	an agent capability
AutoGen (agentchat)	`deploy_service`	the callables you give AssistantAgent
Google ADK	`deploy_service`	the functions in an agent's tools
Claude Agent SDK	`mcp__ops__deploy_service`	PreToolUse hooks
MCP client (Python)	`deploy_service`	any MCP client session
OpenTelemetry	`deploy_service`	your existing tracing
Any other framework	`deploy_service`	any Python callable
TypeScript / JavaScript	`deploy_service`	any async function
HTTP (any language)	`deploy_service`	your own call site

Try CAIN-42 on your own agents

Create a free account and every new account starts with a 7-day trial of the full platform. Or try the sandbox first, with no account at all.

Create a free account → · Try the sandbox · See the whole ecosystem