llms.txt
# Arcjet
Arcjet is the AI agent runtime security platform. Discover the agents running in your organization, enforce policy across every action, prompt, and tool call, and keep the evidence to prove what happened. Detect prompt injection, authorize agent tool calls, redact PII, and block bots and abuse.
Full documentation content: https://docs.arcjet.com/llms-full.txt
Full site index (Learning Center, comparisons, blog): https://arcjet.com/llms-full.txt
Agent registration: https://arcjet.com/auth.md
Arcjet's primary use case is securing the actions AI agents take, when coding on developer laptops or taking actions in production. Agents have gone from answering questions to moving money, changing records, and shipping code, and the security or engineering leader now owns that risk. Identity and RBAC authenticate the agent but don't govern the action it's about to take, and network proxies can't see inside the workflow. Arcjet gives security and engineering teams visibility into what each agent is doing, real-time enforcement before a consequential action (prompt injection, PII, tool authorization), and an audit trail.
Where Arcjet runs: Arcjet enforces policy where the action happens, so there is no network proxy, AI gateway, endpoint agent, or browser extension to deploy. Coding agents call Arcjet from the hooks they already fire, and one policy decides before each tool call runs, in the agent's own process on the developer laptop or cloud agent. Custom agents and web applications call Arcjet from an SDK in their own code. OpenTelemetry and Claude Compliance API ingest add organization-wide visibility, and on the Enterprise plan decisions export to Datadog, Splunk, SentinelOne, Panther, and Amazon S3.
How Arcjet compares to other AI agent security approaches: [AI agent security platforms](https://arcjet.com/compare/ai-agent-security-platforms), [Rein vs Arcjet](https://arcjet.com/compare/rein-vs-arcjet), [Datadog AI Guard vs Arcjet](https://arcjet.com/compare/datadog-ai-guard-vs-arcjet), [AI gateway vs Arcjet](https://arcjet.com/compare/ai-gateway-vs-arcjet), [Runlayer vs Arcjet](https://arcjet.com/compare/runlayer-vs-arcjet).
Arcjet protects several entry points:
- **Coding agents** - apply policies and enforce security controls directly within the development environment. Protect prompts from injection, redact sensitive data, and ensure that agent actions comply with organizational policies. Supported coding agents: Claude Code, GitHub Copilot, Cursor, OpenAI Codex, and Muse Code, enforced through hooks, with OpenTelemetry and the Claude Compliance API for organization-wide visibility. Guide: https://arcjet.com/learn/coding-agent-security
- **Custom agents** - tool calls, queue consumers, agentic pipelines, and anywhere else you process untrusted input in custom agents you've built and deployed yourself. Protect HTTP requests from bots and abuse, and enforce security policies consistently across all agent interactions. Supported AI agent frameworks include: Claude Agent SDK, Claude Managed Agents, Cloudflare Think, CrewAI, Genkit, Google ADK, LangChain, LangGraph, Mastra, OpenAI Agents, Strands Agents, TanStack AI, Vercel AI SDK, Vercel Eve.
- **Web applications** - protect HTTP routes, API endpoints, and middleware within web applications. Bot protection, email validation, WAF, and other security building blocks applied directly in-code. Supported languages and frameworks include: Astro, Bun, Deno, Fastify, NestJS, Next.js, Node.js, Express, Hono, Nuxt, Python, FastAPI, Flask, React Router, Remix, SvelteKit, Go.
Destination threat analysis: a policy can score the URLs, domains, and MCP servers an agent is about to contact (web fetches, URLs inside shell commands, MCP and API calls) with Arcjet threat intelligence and deny high-risk destinations such as malware hosts and rogue MCP servers. It works for coding agents and custom agents. Guide: https://arcjet.com/learn/detect-malicious-urls-ai-agents
## Get started
- [Agent get started](https://docs.arcjet.com/agent-get-started): full agent onboarding flow – install a skill, connect the CLI, add protection.
- [Skills](https://github.com/arcjet/skills): install via `npx skills add arcjet/skills` to give your agent framework-aware integration docs.
- [Arcjet CLI](https://docs.arcjet.com/cli): create sites, retrieve keys, inspect requests, manage rules from the terminal.
- [MCP server](https://docs.arcjet.com/mcp-server): same management surface over MCP at `https://api.arcjet.com/mcp` (OAuth).
- [Arcjet plugin](https://docs.arcjet.com/arcjet-plugin): bundled skills + MCP + coding rules for Claude Code and Cursor.
- [Create an account](https://console.arcjet.com)
- [Quick start guides by framework](https://docs.arcjet.com/get-started)
## Key facts
- Pricing: Individual $25, Startup $299 per month, plus usage. Enterprise is custom priced. 15-day free trial: https://arcjet.com/pricing
- Free plan: after the 15-day trial, accounts continue on a free plan limited to 10,000 requests per month. The limit is a hard stop rather than metered overage.
- SIEM export (Enterprise plan): decisions export to Datadog, Splunk, SentinelOne, Panther, and Amazon S3 for detection, alerting, and long-term retention.
- Security and compliance: SOC 2 Type 2 examination (Security, Availability, Confidentiality); report in the Trust Center: https://trust.arcjet.com
- Cloud API p95 response time around 25ms (goal range 20-30ms): https://arcjet.com/blog/how-we-achieve-our-25ms-p95-response-time-sla
- Company: Arcjet Labs, Inc.; founded by David Mytton (CEO); HQ San Francisco with a New York hub. Seed 2023 (a16z), Series A 2025. Contact: sales@arcjet.com. About page with team, differentiators, and crawlable key facts: https://arcjet.com/about
