Integrations
All brin integrations: Cursor, Claude Code, Codex, Gemini CLI, OpenCode, AI SDK, Mastra, LangChain, AgentMail, Resend. Hooks, wrappers, and middleware.
brin integrates at the tool boundary — the point where your agent touches external context. pick your platform below.
##coding agents
add a hook so brin checks every package before your agent installs it.
| Agent | Hook type |
|---|---|
| Cursor | beforeShellExecution — intercepts shell commands |
| Claude Code | PreToolUse — blocks install commands |
| Codex CLI | Rules — denies direct installs, routes through brin wrapper |
| Gemini CLI | BeforeTool — checks brin before shell commands |
| OpenCode | PreToolUse — gates shell tools |
##agent frameworks
wrap your tools so brin runs before fetch, install, or clone actions execute.
| Framework | Integration point |
|---|---|
| AI SDK | execute function in each tool |
| Mastra | createTool() execution |
| LangChain | DynamicStructuredTool callbacks |
catch phishing in agent inboxes before your agent processes the message.
| Platform | What it does |
|---|---|
| AgentMail | scan incoming agent emails for phishing via webhooks |
| Resend | scan inbound emails before your app processes them |
##enforcement, not guidance
AGENTS.md, system prompts, and policy text are advisory — they don't guarantee the model follows them every time. for real enforcement, put brin at the boundary where the action happens: shell hooks, tool wrappers, or framework middleware. that makes the check part of the actual control flow. if the agent tries to skip it, the tool call still can't proceed.
##don't see your platform?
brin is a single GET request — no SDK, no auth. any tool that can call an HTTP endpoint can check brin before acting. see the API reference and CLI reference.
On this page