# ArmorIQ

> Intent assurance platform that verifies and enforces AI agent plans before actions run, with tamper-evident audit logs.

ArmorIQ is a San Francisco-built control fabric for autonomous AI agents. It captures an agent's declared plan, signs it cryptographically, and checks each action against that signed scope before execution, producing audit logs of every decision.

## What It Is

ArmorIQ is an agent security and governance layer that sits between AI reasoning and execution. Where IAM controls who can access a resource, ArmorIQ controls what an agent does once it has access. Agents plan and adapt freely, but any action outside the approved task is blocked. It is meant to complement existing IAM, Zero Trust, and monitoring rather than replace them.

## How Intent Assurance Works

The workflow has three steps: discover, enforce, and prove. Agents and MCP servers are added to a registry with ownership and org-level policy. Each agent declares its plan, which ArmorIQ captures and signs, then policy and runtime checks verify every action against that signed plan. Allowed actions continue to MCP servers and downstream tools. Blocked actions do not execute, and the agent receives an explanation such as 'Action blocked: exceeds delegated authority.' Every allow, block, or escalate decision is sealed into a tamper-evident audit log, which the site says maps to SOC 2, GDPR, and NIST.

## Platform Components

The site describes five layers: Intent Engine (policy decisions, which the site says take under 5ms), Sentry (real-time action visibility and alerts), Gatekeeper (resource access control for AI behavior), Registry (a dashboard of deployed agents), and Auditor (compliance-ready logs). It also lists an Intent Container and agent-specific tools: ArmorClaude, ArmorCodex, ArmorOpenCode, ArmorCopilot, ArmorGemini, and ArmorClaw.

## Deployment and Licensing

ArmorClaw, intent assurance for OpenClaw agents, is open source. The full ArmorIQ control fabric is a commercial product with self-hosted and cloud deployment options. The vendor states it is built for security teams, platform engineers, and compliance officers running agents in production, and cites figures such as 44K+ SDK installs and 9K+ agents live in production.

## Features
- Cryptographically signed agent intent plans
- Runtime enforcement of actions against signed scope
- Agent and MCP server registry with ownership and org-level policy
- Real-time action monitoring and alerts (Sentry)
- Resource access control for AI behavior (Gatekeeper)
- Tamper-evident audit logs mapped to SOC 2, GDPR, NIST
- Blocking of out-of-scope actions with explanations
- Self-hosted and cloud deployment options

## Integrations
MCP servers, Slack, GitHub, Gmail, Notion, OpenClaw, Claude, Codex, OpenCode, GitHub Copilot, Gemini

## Platforms
LINUX, WEB, API, DEVELOPER_SDK, CLI

## Pricing
Paid

## Links
- Website: https://armoriq.ai
- Documentation: https://docs.armoriq.ai/
- Repository: https://github.com/armoriq
- EveryDev.ai: https://www.everydev.ai/tools/armoriq