Ship AI Agents That Act. Govern Every Action.
Shrike governs the action an agent takes, not just the prompt it's given, applying your policy in real time to what's allowed, blocked, or escalated. Shrike doesn't decide on its own authority; you declare the policy and Shrike enforces it, at a point the agent can't reason its way around. That runtime control is what moves AI agents from proof-of-concept to production. Without it, they stall in security review. With Shrike, they ship.
How it works
Your agent, or the framework it runs in, sends Shrike the action it is about to take: a shell command, a file write, a SQL query, a web request, a tool call over MCP, a message to another agent. Shrike checks it against the scope you declared for that agent and the policies you turned on, and answers before the action runs: allow, redact, block, or hold for a person. Deterministic pattern checks answer in under 15ms; deep semantic analysis follows in seconds.
A refusal is not a dead end. It tells the agent why, and what would clear it, so the agent can change the action instead of retrying it. An approval policy can name an action class, such as bulk exports, so a person decides on the dashboard while the session keeps going. And an operator can stop an agent: every governed action is refused, with the reason, until a person releases it. Actions already in flight and integrations that never call Shrike are outside it.
What the operator sees
A dashboard with every agent and its declared scope, and a session view that reads turn by turn: what was said, what was done, what was held or stopped, and who decided it. Policy templates for SOC 2, HIPAA, GDPR, PCI DSS and ISO 27001 become enforcing rules when you activate them.
Integrations
Six channels: REST API, MCP server (15 tools, works with Claude Code and any MCP client), LLM gateway, and SDKs for Python, TypeScript and Go. The SDKs ship adapters for LangChain and LangGraph, the OpenAI Agents SDK, Google ADK, the Claude Agent SDK and the Vercel AI SDK. A Claude Code plugin governs a coding agent's commands and file writes out of the box. The MCP server and the SDKs are open source (Apache-2.0) at github.com/Shrike-Security.
Pricing
Start free. The Community tier runs the deterministic checks with a dashboard and scan history, 1K scans a month. Pro and Team add the semantic layers and higher limits. Enterprise is custom, with a private-cloud kit that can be deployed in your own environment.
Why we built it
Prompt scanners stop bad inputs. Shrike stops bad inputs and the bad actions they can't see. Agents stall in security review because nobody can say, in an enforceable way, what they are allowed to do. Shrike turns that into a declared policy, enforced at runtime, with a record of every decision.
Built with