Policy outside prompts
Define allow, deny, and approval-required actions outside the model prompt.
RuleOak Core v2.0.1
RuleOak wraps agent actions with policy checks, evidence records, approval gates, and audit reports — without forcing developers to redesign their app.
npm install
npm run integrate:10min
npm run approval:inbox:build
npm run report:htmlThe shortest path: declare a tool, evaluate policy before execution, record evidence, pause for approval when required, then export an audit report.
Developer value
Keep your agent framework, your tools, and your local workflow. RuleOak sits at the action boundary where safety and accountability matter.
Define allow, deny, and approval-required actions outside the model prompt.
Pause external messages, writes, and high-risk tool calls before execution.
Attach decision context so recommendations and actions are reviewable.
Generate local reports for tool decisions, approvals, and run history.
Start here
The 10-minute demo is the recommended first experience for developers evaluating RuleOak.
npm install
npm run integrate:10min
# search_docs: allowed
# send_external_message: approval_required
# delete_workspace_file: blockedPlatform shape
Wrap any proposed tool call with policy, evidence, approval, and audit.
Prototype governance boundary for MCP-style tool calls and proxy patterns.
Review approval-required actions locally before anything risky proceeds.
Interoperability
RuleOak Core uses a versioned governance record protocol so SDKs, adapters, reports, and future connectors can share the same accountability model.
Schemas and conformance checks for core governance records.
LangGraph-style and CrewAI-style samples show how RuleOak can wrap existing agent flows.
Private-preview Python SDK guidance for emitting compatible governance records.