Documentation

Everything you need to secure your AI agentsFrom a two-minute install to org-wide policy, telemetry, and the SDKs for agents you build yourself.

Quickstart

Install, run the setup wizard, enroll your device, and hook your agent. Five minutes, step by step.

Let your agent set it up

Paste one skill into Claude Code, Cursor, or any agent. It installs the CLI, wires the hooks, and verifies everything.

Prismor Overview

Open-source security for AI coding agents. Prismor intercepts every tool call before execution and blocks the dangerous ones.

Connect to the Platform

Enroll devices, stream redacted telemetry, and push signed org policy from the control plane to every machine.

Runtime & Policy

Authoring a Policy

Detection rules, custom patterns, and per-project overrides in policy.yaml.

Layered Policy & Exemptions

Org floor, project overlays, and time-boxed exemptions. Decide who can relax what.

Architecture

How the hook dispatcher, policy engine, and telemetry pipeline fit together.

Secret Protection

Cloak

Secrets become placeholders before the model ever sees them. Real values never enter context.

Sweep

Find and redact secrets already leaked into AI tool caches and configs.

Hermes Cloaking

Entry-point cloaking and paste guard for the Hermes agent.

Detection & Response

Semantic Guard

LLM-backed intent check that catches prompt injection and paraphrased attacks.

Canary

Honeytokens that trip the moment an agent goes looking for credentials.

Scoped Agent

Pin a session to its stated task and flag pivots outside it.

Learning

Session-history rules that cut false positives and catch evasion.

Identity & Observability

IAM

Least-privilege identity profiles for agents.

Live Telemetry

Device enrollment, redacted event streaming, and signed policy pull.

Dashboard

Local terminal + web dashboards for sessions and forensics.

Framework Adapters

Adapters Overview

Guard agents you build, not just agents you run. Coverage matrix and patterns.

LangChain / LangGraph

guard_tools() wraps every tool with policy checks.

CrewAI

Guard crews and their tools per agent.

OpenAI Agents SDK

Function-calling guardrails for the Agents SDK.

Browser Use

Policy checks on every browser action the controller takes.

Vercel AI SDK

TypeScript adapter. prismorTools() + useSubject(), LangGraph JS, and the eval-server sidecar.

SDK Integration

The HTTP adapter pattern for any language, with per-user subjects.

Supply Chain & Network

Supply Chain

Risk-scores npm / pip / cargo installs against age, maintainers, and IOC feeds.

Network Isolation

Egress allowlists, raw-IP blocking, and tunnel detection for agent traffic.

Skill Scanner

Scans MCP servers and agent skills for risky capabilities.

Integrations & Deploy

GitHub App

PR vulnerability scanning and AI-powered fixes on every pull request.

CLI Documentation

Install, authenticate, and scan repositories locally or in CI.

Build Scanning

GitHub Actions and CI pipeline security checks.

Auto-Fix

AI-generated pull requests that patch what the scans find.

Docker & Containers

Container prerequisites and sandbox hardening.