raigo cloud

raigo is an open standard.The core of this repository is the.raigofile format — a YAML-based, declarative policy definition language for AI governance. It is the single source of truth for an organisation's AI usage rules: what is allowed, what is blocked, what must be logged, and which compliance frameworks apply.

The.raigoformat is the open-source contribution. It is free to use, implement, extend, and build on. Anyone can write a.raigofile, build a compiler that reads it, or build an engine that enforces it. The format is not tied to any vendor, platform, or cloud product.

The.raigoformat is to AI governance what OpenAPI is to REST APIs, or what Rego is to policy-as-code: a vendor-neutral, open specification that any tool can implement.

A.raigofile is a YAML document. It has three sections:metadata(who owns the policy, compliance context),context(the environment: tools, data classifications, networks), andpolicies(the rules: atomic, one directive per rule, deterministic action).

A.raigofile can be used in two ways, depending on your deployment model.

Compiler mode— for tools that cannot call an external engine (ChatGPT, Claude.ai, n8n, Lovable, GitHub Copilot). The raigo compiler reads the.raigofile and generates a native enforcement artifact for the target platform: a system prompt, a JSON config, a Markdown knowledge block. The AI platform enforces the policy using its own native mechanisms.

Engine mode— for applications that can make HTTP calls. The raigo engine runs as a lightweight service. Before a prompt reaches the LLM, your application callsPOST /v1/evaluate. The engine evaluates the prompt against the active policy and returns a deterministicALLOW,DENY, orWARNdecision. Nothing reaches the LLM until the engine approves it.

Your Application │ ▼ ┌─────────────┐ ┌──────────────┐ │ raigo engine│─ ALLOW ─▶ LLM API │ │ (policy. │ │ (OpenAI, │ │ raigo) │─ DENY ─▶ Anthropic, │ └─────────────┘ │ │ etc.) │ │ └──────────────┘ ▼ Violation Response (action, policyMessage, matchedRules, auditLog)

The key difference from system prompts is that enforcement in engine mode isdeterministic. ADENYrule cannot be overridden by prompt injection, model drift, or a creative user. The engine blocks the request before the LLM ever sees it.

The same.raigopolicy file works across all four deployment models without modification.

When running in compiler mode, raigo generates native enforcement artifacts for each target platform. Each output includes runtime handler instructions telling the platform how to enforce the policy.

# Install the compiler npm install -g @periculo/raigo # Compile for a specific platform raigo compile policy.raigo --target claude raigo compile policy.raigo --target n8n # Compile for all platforms at once raigo compile policy.raigo --all # Validate a policy file against the spec raigo validate policy.raigo # Initialise a new policy from a template raigo init

When you first deploy the raigo engine, it starts inObserve modeby default. In this mode, allDENYrules are downgraded toWARN— nothing is blocked, but every policy match is logged. This lets you see exactly what your rules would have blocked across your real traffic before committing to enforcement.

Once you are confident your rules are correct, switch toEnforce mode— at which pointDENYrules become active.

# Start in observe mode (default) RAIGO_ENGINE_MODE=observe raigo serve --policy master.raigo # Switch to enforce when ready RAIGO_ENGINE_MODE=enforce raigo serve --policy master.raigo

If you are building an AI tool, agent framework, or workflow platform, you can add native raigo support so your users can govern it with a.raigofile.

Option 1 — Engine integration.Configure your platform to call the raigo engine API (POST /v1/evaluate) before executing AI actions. Your platform passes the proposed prompt to raigo and receives a deterministicALLOW/DENY/WARNdecision. This is the recommended approach for any platform that can make HTTP calls.

Option 2 — Compiler integration.Add a raigo compilation target for your platform's native policy format. When users runraigo compile policy.raigo --target yourplatform, raigo generates the correct artifact for your platform to load and enforce natively.

Join the raigo Discord to discuss the specification, share.raigopolicies, get help with integrations, and follow the roadmap.

The most impactful contributions are to thespecification itself: new field definitions, new compliance framework mappings, new condition trigger types, and new compilation targets. If you are implementing raigo in a new platform or tool, please open an issue so we can link to your implementation.

ThePericulo Enterprise Control Planeprovides centralised policy management, automatic sync to connected tools, fleet-wide enforcement, and a full audit trail across all deployment models.

Manage Akamai's edge platform, including properties, DNS, certificates, security, and performance optimization, using AI assistants.

Provides a unified interface to AWS services for security investigations and incident response.

An MCP server that enables AI assistants to interact with AWS security services.

DevOps MCP — Secure MCP Server for Linux Server Automation

A three-tier access control MCP server that allows AI assistants (Claude Code, Cursor, Windsurf) to safely scan, plan, and operate Linux servers via SSH without full write access. Includes an out-of-band human consent token gate, automated port-conflict scanning, and a completely read-only default safe mode to eliminate accidental destructive commands on production environments.

An MCP server for Alibaba Cloud's Edge Security Acceleration (ESA) service.

Securely manage secrets and policies in HashiCorp Vault through an MCP interface.

Interact with the Illumio Policy Compute Engine (PCE) to manage workloads, labels, and analyze traffic flows.

Interact with Jamf Pro for Apple device management tasks.

Interact with JupiterOne's data and tools through an MCP server, enabling AI assistants to access your JupiterOne account.