Impact-Site-Verification: ea5f7f08-74ae-4cca-af2a-01826f2587ea
// inline guardrails for claude code, copilot, codex, gemini cli, opencode
The missing security layer for AI agents
Real-time protection against prompt injection, data leaks, and destructive commands. Enable or disable individual command, web, and LLM checks from the dashboard, or start from preset allow lists.
drops into the agents you already run
Defense in depth for every prompt
Layered checks, run in parallel. Each request passes through multiple checks at once and short-circuits the moment a real threat is found — fast processing, low-latency replies.
You're in full control. Enable, disable, or customize checks easily from your dashboard. AgentGuards is fully hosted on secure AWS infrastructure.
Your data stays yours. We don't store your prompt content — only pass/fail results, used to improve detection.
Prompt injection & jailbreak
Heuristic patterns, an optional LLM judge, and Meta's PromptGuard ML model catch instruction-override and jailbreak attempts before they reach your model.
PII & secret redaction
Detect and redact emails, SSNs, credit cards, API keys, and tokens in prompts and tool inputs.
Data exfiltration & restricted topics
Block attempts to siphon data or steer the agent into off-limits territory, with policy-as-code you control.
Credential egress blocking
When an agent runs an outbound command, we decode base64/hex/URL payloads and stop credentials — API keys, tokens, private keys — from leaving in a request body or header, no matter the destination.
Web-content defense
Purpose-built checks for fetched web pages — hidden text, HTML-comment injection, and XSS-as-injection vectors.
Code scanning (SAST + secrets)
Scan the code an agent writes for vulnerabilities and hardcoded secrets — Semgrep static analysis plus gitleaks — before it lands.
Per-tenant config & isolation
Every customer gets isolated checks, their own API keys, and full dashboard control over which checks run.
Bring-your-own model key
Run the LLM-judge check on your own OpenAI key — your traffic, your spend, full control.
Live footage, not a mockup
Real attacks, caught at the door.
Prompt-injection probes, instruction overrides and destructive commands get scored and stopped before your agent ever acts on them.
Watch AI Attacks Get Blocked in Real Time →Frequently asked questions
Where is the service hosted?+
Our frontend and backend are deployed on AWS (eu-north-1 region). All traffic is served over HTTPS via an Application Load Balancer with ACM certificates.
How do I integrate AgentGuards?+
The quickest path is our MCP server — one command wires it into Claude Code, VS Code Copilot, or Codex. You can also call the REST API directly. See the integration guide →
What checks does AgentGuards run?+
Every prompt goes through up to 10 layered checks: prompt injection, jailbreak detection, PII detection, secret detection, data exfiltration, toxicity, restricted topics, web-content injection, an LLM semantic check, and a PromptGuard ML classifier — checks short-circuit on the first confirmed threat. On top of that, AgentGuards scores every shell command your agent tries to run and blocks destructive ones (like rm -rf /), and scans every file it writes or edits for hardcoded secrets and vulnerable code (SAST via semgrep + gitleaks).
Does AgentGuards store my prompts?+
No. We scan prompt content in memory and discard it immediately. Only metadata — request counts, token counts, blocked event types — is persisted for your usage dashboard.
Which LLMs and frameworks are supported?+
AgentGuards is model-agnostic. It sits in front of any LLM call as an HTTP guardrail. Native integrations exist for Claude Code, Gemini CLI, OpenAI Codex, and VS Code GitHub Copilot.
What happens when a threat is detected?+
The request is blocked before it reaches the model. You receive a JSON response with decision: "block" and a per-check breakdown showing which check triggered and why. Your LLM is never called.
Can I configure which checks run?+
Yes. Every check can be toggled on or off per tenant from your dashboard. Individual and higher plans can also edit custom detection patterns.
What counts as a request?+
Each guardrail evaluation — an input check, output validation, action authorization, policy evaluation, or gateway completion — is one metered request.
Do AgentGuards requests use up my Claude Code (or other LLM) credits?+
Depends on which integration you use. With the hook-basedintegration (Claude Code, Gemini CLI, Codex hooks), our check runs as a plain HTTP call outside the model's context — Claude never sees it, so it burns zero of your model credits. With the MCP serverintegration, calling our tools costs a small amount of Claude context/tokens, the same as any MCP tool call, since the model has to load the tool schema and read the JSON result — that's what shows up under "MCP servers" in Claude Code's /usagebreakdown. Either way, the security check itself never invokes an LLM on our end (it's regex plus a self-hosted classifier), and it only counts once against your AgentGuards request quota — a separate meter from your LLM credits. The one exception is the optional LLM-judge check, which — if you enable it — runs on your own OpenAI key, so any token cost there stays on your account, not ours.
Is the free plan really free? Do I need a credit card?+
Yes. The Free plan is $0, forever — 1,000 requests per month, 1 API key, no credit card required to start. It's not a time-limited trial: if you never upgrade, you keep that quota, reset every month, for as long as you use AgentGuards.
What happens when I hit my monthly limit?+
Requests beyond your plan's included quota are paused until the next cycle or an upgrade. You'll see usage and remaining quota in your dashboard, and we warn you before you run out.
Can I bring my own model key?+
Yes. The optional LLM-judge check runs on your own OpenAI key, so that spend stays on your account.
Do you offer on-prem?+
Enterprise can deploy in your own VPC or on-prem, with SLA, SSO, and a security review.
Ship your agent with guardrails on.
Start free in two clicks. No credit card, no sales call.