AgentGuards - Self-Hosted LLM Guardrails and AI Governance
Agentguards · Cybersecurity & IT
Certification per AWS Marketplace.
Evidence tier Source Confirmed · 4 captures on record
What the publisher says
As described on AWS Marketplace.
AgentGuards sits between your applications and any LLM provider and enforces policy on the way in and the way out - as a single appliance running in your own AWS account.
Input screening: Every request is checked for prompt injection, jailbreak attempts, data-exfiltration patterns, personal data, credentials and secrets, toxicity, and topics you have placed off-limits. Alongside the transparent pattern-based checks, the image bundles a fine-tuned prompt-injection classifier that runs locally on CPU - no GPU, no external inference service.
Show the rest of the publisher’s description (4 more lines)
Output validation: Model responses are checked before they reach your users, for leaked personal data or credentials, schema conformance, citation and grounding requirements, and your own content policy.
Agent action governance: Shell commands an agent proposes are risk-scored before they run, so a destructive or credential-reading command can be blocked or held for approval rather than discovered afterwards.
A governance console, not just a filter: Every decision is recorded with a deterministic risk tier and mapped to OWASP LLM Top 10 and MITRE ATLAS categories. Administrators get role-based access with optional SSO, a review queue for high-risk refusals, a prompt tester for validating rules before deployment, retention controls, and an inventory that discovers which models and tools are actually in use - including the ones nobody told you about.
Deployed as an appliance you control: Launch it in a private subnet, reach it over HTTPS on your own network, and point your agents at it. It generates its own credentials on first boot - no secret ships in the image - and every check runs on the instance. The only outbound connection AgentGuards requires is AWS License Manager, to verify your subscription.
Highlights
Highlighted by the publisher on AWS Marketplace.
Runs entirely in your VPC: Prompts, completions and code never reach AgentGuards or any third party. The detection model ships inside the AMI - there is no callback, no telemetry of your content, and it works with no internet egress beyond AWS licence verification.
Governance, not just filtering: Role-based access (owner / operator / viewer), a deterministic risk tier on every decision, a searchable audit trail, a human review queue, and a discovered inventory of every model and tool your agents actually use.
Works with the agents your developers already run: Drop-in integrations for Claude Code, Codex, Gemini CLI, GitHub Copilot CLI and OpenCode, plus an OpenAI-compatible gateway, an Anthropic proxy and a plain REST API.
Agent build and provenance
See the full provenance
The layer-by-layer build, the evidence behind each claim, the risk basis and the cross-marketplace links are open to any account. Some rows are disclosed, some the source leaves Unknown; a free account shows you which.
Plans and pricing as listed
8 listed- Hrs
- Hrs
- Hrs
- Hrs
- Hrs
- Hrs
- Hrs
- Hrs
Refund terms
As stated by the publisher on AWS Marketplace.
Fees are non-refundable. You may cancel at any time; cancellation stops future charges, but amounts already paid or committed for the current License Term are not refunded. To request a refund or ask a billing question, contact support@agentguards.co
Sources
Linked repositories
Unknown means this listing does not publish a repository. It is not a statement that the code is closed, and a linked repository is not a claim that the publisher wrote it: the registry computes that relationship privately and does not publish it.
Evidence risk is the share of the build you cannot see before you deploy, not a security rating. Sign in to see the layer-by-layer basis for this band.