Sold by: Agentguards
Security guardrails for AI agents and LLM applications.
AgentGuards screens every prompt, tool call and model output for prompt injection, jailbreaks, PII and secret leakage, and data exfiltration then governs
the actions your agents take, blocking destructive commands before they run.
Every decision is written to an auditable trail tagged to OWASP LLM and MITRE ATLAS.
Integrates in minutes with Claude Code, OpenAI Codex, Gemini CLI and GitHub Copilot, OpenCode or directly via REST and an OpenAI-compatible gateway.
Overview
AgentGuards is a security layer for AI agents and LLM applications. It screens every prompt, tool call and model output before it reaches your model - and governs the actions your agents take afterwards.
THE PROBLEM An AI agent with real permissions is an insider that never sleeps and never questions its instructions. Prompt injection is not a content-moderation problem in that setting; it is privilege escalation by another name. A single poisoned web page, ticket or file can redirect an agent that has your credentials, your repositories and your production access.
WHAT AGENTGUARDS DOES Input screening. Every prompt and tool input is checked for prompt injection, jailbreak attempts, PII, hardcoded secrets, and restricted topics - combining fast deterministic patterns with a fine-tuned ML classifier, so novel phrasings are caught rather than only known strings.
Output and egress control. Model outputs and outbound requests are screened for data exfiltration. Base64, hex and URL-encoded payloads are decoded first, so credentials cannot leave in an encoded request body or header regardless of destination. Action governance. Before an agent runs a shell command, deploys, or deletes, the action is authorised against policy. Destructive commands are blocked or escalated for human approval rather than executed and regretted.
Code and dependency scanning. Generated code is scanned for vulnerabilities and hardcoded credentials, and dependencies are checked against known CVEs, before anything is committed or run.
AUDITABILITY Every allow, block and redaction is written to a decision trail tagged to the OWASP LLM Top 10 and MITRE ATLAS, with the policy version that produced it. Retention is configurable, and raw prompt content is deliberately not stored in the audit trail.
INTEGRATIONS Drop-in support for Claude Code, OpenAI Codex, Gemini CLI and GitHub Copilot via hooks and plugins - no code changes to your application. Or call the REST API directly, use the OpenAI-compatible gateway as a drop-in base URL, or connect over MCP.
WHAT IS INCLUDED The Team Max annual contract includes 750,000 screened requests per month, 25 API keys, every guardrail and scanner listed above, a dashboard with per-check configuration and a prompt tester, and priority email support.
GETTING STARTED After purchase, choose Set up your account in the AWS Marketplace console. You will be brought to agentguards.co to create your account, and your subscription is linked automatically. Most teams are screening live traffic within ten minutes.
Highlights
- Stop attacks before they reach your model. AgentGuards screens every prompt, tool call and model output for prompt injection, jailbreaks, PII and secret leakage, and data exfiltration, combining fast pattern checks with a fine-tuned ML classifier, so novel phrasings are caught, not just known strings. Blocked requests return a clear reason your developers can act on.
- Govern what your AI agents do, not just what they say. Agentic AI has real permissions, which makes prompt injection a privilege-escalation problem. AgentGuards authorises the actions agents take, blocking destructive commands, scanning generated code for vulnerabilities and hardcoded secrets, and checking dependencies for known CVEs before anything runs.
- Prove it to your auditors. Every allow, block and redaction is written to an auditable decision trail tagged to OWASP LLM Top 10 and MITRE ATLAS, with per-tenant policy versioning and configurable retention. Integrates in minutes with Claude Code, OpenAI Codex, Gemini CLI and GitHub Copilot, or directly via REST and an OpenAI-compatible gateway.
Details
New
Introducing multi-product solutions
You can now purchase comprehensive solutions tailored to use cases and industries.
Features and programs
Financing for AWS Marketplace purchases
AWS Marketplace now accepts line of credit payments through the PNC Vendor Finance program. This program is available to select AWS customers in the US, excluding NV, NC, ND, TN, & VT.
Pricing
Pricing is based on the duration and terms of your contract with the vendor. This entitles you to a specified quantity of use for the contract duration. If you choose not to renew or replace your contract before it ends, access to these entitlements will expire.
Additional AWS infrastructure costs may apply. Use the AWS Pricing Calculator to estimate your infrastructure costs.
Dimension | Description | Cost/12 months |
|---|---|---|
Team Max (annual) | Annual contract for the AgentGuards Team Max plan. Includes 750,000 screened requests per month, 25 API keys, and every guardrail: prompt injection and jailbreak detection, PII and secret redaction, data exfiltration blocking, source-code and dependency scanning, and action governance for destructive commands. Every decision is written to an auditable trail tagged to OWASP LLM Top 10 and MITRE ATLAS. Includes priority email support. Your subscription is activated by signing in at agentguards.co after purchase | $2,988.00 |
Dimensions summary
You buy a single annual contract for the Team Max plan. Pricing is one fixed yearly commitment, not usage-based or per-seat billing. Your contract covers 750,000 screened requests per month and 25 API keys. Every guardrail is included in this one plan, so there are no add-ons or feature tiers to select. When you deploy the self-hosted appliance in your own AWS account, the price stays the same across supported instance sizes. You size the instance for your traffic without changing the software fee. AWS bills you directly and manages the subscription.
Top-of-mind questions for buyers
Each guardrail evaluation is one request. That includes an input check, output validation, action authorization, policy evaluation, or gateway completion. The security check itself never calls an outside model, so it counts once against your quota regardless of which integration you use.
Requests beyond your monthly quota pause until the next cycle begins. You see your usage and remaining quota in your dashboard, and you get a warning before you run out. The quota resets each month rather than carrying unused requests forward.
No. The software fee stays the same across all supported instance sizes, so you size the instance for your traffic, not your budget. You still pay the underlying AWS EC2 compute cost for whatever instance you choose to run.
www.agentguards.co+1
Helpful?
Vendor refund policy
Annual contracts are non-refundable once the term begins; unused time is not refunded. Stop renewal any time in the AWS Marketplace console. At our discretion we refund a billing error (including being charged twice for one period) or an AgentGuards outage over 24 consecutive hours reported within 7 days. Approved refunds are processed through AWS Marketplace. Contact support@agentguards.c
Legal
Vendor terms and conditions
Upon subscribing to this product, you must acknowledge and agree to the terms and conditions outlined in the vendor's End User License Agreement (EULA) .
Content disclaimer
Vendors are responsible for their product descriptions and other product content. AWS does not warrant that vendors' product descriptions or other product content are accurate, complete, reliable, current, or error-free.
Delivery details
Software as a Service (SaaS)
SaaS delivers cloud-based software applications directly to customers over the internet. You can access these applications through a subscription model. You will pay recurring monthly usage fees through your AWS bill, while AWS handles deployment and infrastructure management, ensuring scalability, reliability, and seamless integration with other AWS services.
Support
Vendor support
Email: support@agentguards.co responses within one business day. Documentation is built into the console at /admin/ui/docs, with a support page covering diagnostics and admin lockout recovery.
AWS infrastructure support
AWS Support is a one-on-one, fast-response support channel that is staffed 24x7x365 with experienced and technical support engineers. The service helps customers of all sizes and technical abilities to successfully utilize the products and features provided by Amazon Web Services.
Similar products
Screen prompts and model output for injection, jailbreaks, PII and secrets inside your own VPC. Includes a governance console with role-based access, risk tiers and an auditable decision trail. Nothing you send it ever leaves your network.
CyberArk Agent Guard is an AI agent security tool.
Agent Guard can be used to secure secrets retrieval for agents managed via an external secret provider, such as AWS Secrets Manager and CyberArk Secrets Manager, and traceability of AI agent MCP communications via Agent Guard MCP Proxy.
DataRobot's Agent Workforce Platform lets enterprises build, operate, and govern fleets of production AI agents on AWS - and in hybrid, sovereign, and air-gapped environments - with the same governance layer in every one. Model-, framework-, and orchestration-neutral, with forward-deployed DataRobot engineers who build your first mission-critical agents alongside your team.
Intelliguards MDR defends your organization with high confidence using the most advanced technologies, threat intelligence, and expertise. The service provides continuous monitoring, triage, and advanced response.
Customer reviews
No customer reviews yet
Be the first to review this product . We've partnered with PeerSpot to gather customer feedback. You can share your experience by writing or recording a review, or scheduling a call with a PeerSpot analyst.