# AI-FW Gateway Review Self-hosted AI firewall and governance gateway from Securetron that inspects, filters, and governs every agent and model interaction in ## Editor's Score: 72/100 Capability scores well because AI-FW addresses a genuinely underserved problem: deterministic, fail-closed enforcement at the gateway layer with cryptographic agent identity, three-engine semantic analysis, and GRC automation in a single self-hosted container. Few tools in this directory combine all of these. Ease of use loses points because self-hosting a PostgreSQL-backed container with mTLS, SCIM, and risk profile configuration is not a low-friction setup, and the documentation quality could not be independently verified from available sources. Value is solid for the Starter tier, which gives the full platform including SSO and SCIM at zero cost with 100 million tokens per month, but the complete absence of any published Enterprise pricing or mid-tier option is a real friction point for budget-constrained teams. Delivery scores below average because no independent user reviews or third-party validation were findable, the Capterra listing returned only generic placeholder content, and the product appears to be very recently launched based on the press release date, meaning its track record in production environments is unproven. ## Pros - Fail-closed by design: a scanner error or missing configuration blocks the request rather than silently forwarding it, which is the correct default for a security enforcement layer. - Starter tier includes the full platform, including OIDC SSO, SCIM provisioning, semantic intent analysis across all three engines, and 100 million tokens per month, all at no cost and with no credit card required. - Metadata-only audit logging means raw prompt and response content is never persisted, which directly addresses data-residency and privacy requirements without requiring custom log scrubbing. ## Cons - Enterprise pricing is entirely opaque: no published price, no starting figure, no per-token or per-seat rate on the public page. Teams that need a budget number before engaging sales are blocked from even beginning an evaluation. - The Starter tier's 100 million token-per-month ceiling sounds large but can be consumed quickly by agentic workloads running multiple model calls per task; there is no published overage rate or mid-tier option between free and custom Enterprise. - M365 Copilot admin sync and mTLS/Kerberos agent authentication are gated exclusively to Enterprise, meaning any organization that needs Microsoft 365 integration must commit to a custom contract before they can test that specific capability. ## Pricing Freemium. A free tier is available. - Starter: Free - Enterprise: Custom Pricing reflects what we saw at the time of review (2026-10). Confirm current pricing on the official site. Category: AI Agents & Automation (https://aitoolseekers.com/category/ai-agents-automation) Official site: https://aifw.io/ Last verified by a human: 2026-10-18 Reviewed by a human at AI Tool Seekers, a hand-reviewed AI tools directory. Every listed tool is verified by a person before publishing. Scores are editorial opinions from a fixed rubric (capability, ease of use, value, delivery on promise) and cannot be bought; there is no pay-to-rank. Directory: https://aitoolseekers.com This review: https://aitoolseekers.com/tools/ai-fw-gateway Methodology: https://aitoolseekers.com/how-verification-works ## Verdict AI-FW Gateway is the right pick for enterprise security teams that need cryptographic agent identity, deterministic policy enforcement, and full audit trails without sending prompt data to a third-party cloud. Its Enterprise tier is custom-priced with no published figures, which makes budget planning opaque for teams that need a number before they can get approval. Teams that want a simpler, code-first AI workflow layer without the security-governance overhead should look at TRam Studio instead. Best for: Enterprise security and compliance teams governing AI agent fleets on-premises Not for: Small dev teams wanting a lightweight, code-first LLM proxy without governance overhead ## Overview AI-FW Gateway is a self-hosted AI firewall built by Securetron, the same company behind PKI Trust Manager and Boltron-AI. Its core premise is that every prompt going to a model and every response coming back should pass through a deterministic enforcement layer that you control, running in your own environment, before anything reaches a third-party provider. That is a meaningfully different position from cloud-hosted AI gateways like Cloudflare's or Vercel's, where your traffic transits someone else's infrastructure. The product is organized around several distinct security functions: agent identity and authentication (cryptographically verifiable identities per agent, workflow, and model invocation), semantic intent analysis across three engines for prompt injection and jailbreak detection, PII masking before data leaves the secure perimeter, conditional routing across more than 200 LLMs with failover and load balancing, and a completion cache that handles both exact and semantic matches. GRC automation sits on top of all of this, generating audit evidence and compliance mapping automatically. Deployment is a single container backed by PostgreSQL, and any OpenAI-compatible or Anthropic-compatible client can point at the gateway with only a base-URL change. The fail-closed default is a deliberate design choice: a scanner error or missing configuration blocks the request rather than forwarding it unfiltered. That is the right call for a security product, and it distinguishes AI-FW from lighter-weight proxies that default to pass-through on error. The product holds ISO/IEC 27001, ISO/IEC 42001, and SOC 2 Type II certifications, which matters for enterprise procurement. The Starter tier is genuinely free at 100 million tokens per month with no credit card required, covering the full platform including OIDC SSO and SCIM provisioning. Enterprise pricing is custom and requires a sales conversation. ## Who should use it AI-FW Gateway is built for enterprise security engineers, compliance officers, and platform teams that need to govern AI agent behavior at the infrastructure level, not the application level. If your organization runs multiple AI agents calling multiple models and you need cryptographic proof of which agent called what, when, with what policy applied, this is one of the few tools that addresses that problem directly. It is also a strong fit for regulated industries where raw prompt content must never leave the corporate perimeter: the metadata-only logging model means sensitive data stays out of the audit store entirely. Teams that are still in early AI exploration, running a handful of API calls from a single application, will find the governance machinery here heavier than they need. For workflow automation with built-in auditability on Azure infrastructure, TRam Studio is a better-matched option. For teams that simply want to coordinate specialist AI agents on complex tasks without the security layer, Orkas offers a local-first, open-source alternative that does not require standing up a gateway. ## FAQ Q: How is AI-FW Gateway deployed? A: It runs entirely in your own environment, on-premises or in your private cloud, as a single container service backed by PostgreSQL. Your traffic never needs to leave your network. Q: Does the free Starter tier include SSO and SCIM? A: Yes. OIDC SSO and SCIM provisioning are included in the free Starter tier, which covers 100 million tokens per month with no credit card required. Q: What models does AI-FW Gateway support? A: Any OpenAI-compatible or Anthropic-compatible backend, including hosted providers such as OpenAI, DeepSeek, Anthropic, Mistral, Groq, and OpenRouter, plus self-hosted engines like Ollama, vLLM, and LM Studio. The routing layer covers more than 200 LLMs. Q: Does AI-FW store raw prompt content? A: No. The platform is metadata-only: it logs models, identities, latency, and decisions with truncated previews. Raw prompt and response content is never persisted. Q: What happens if the gateway encounters an error? A: AI-FW is fail-closed by design. A scanner error or missing configuration blocks the request rather than forwarding it unfiltered to the model. Q: How much does the Enterprise tier cost? A: Enterprise pricing is custom and tailored to your deployment. No published price exists; you need to contact the sales team to get a figure.