What AgentFort is
AgentFort is security for AI agents, built by Sec-Astra, a US company. It finds every agent, gives each one an identity and an owner, limits each call to what the job needs, stops sensitive data leaving, contains a manipulated agent, and keeps a receipt for every request. Our aim fits in one line: put AI agents to work without putting your business at risk.
Despite the similar name, AgentFort is not affiliated with Salesforce Agentforce.
The problem AgentFort solves
AI agents no longer just answer questions. They do real work inside companies, and they often get far more access than their job requires. Sensitive data can slip out through them. A hidden instruction in a document can trick them into actions nobody approved. And security teams often can't say which agents are running, or what they did.
Most controls answer one question: can this agent reach this system? AgentFort checks the call itself: the operation, its arguments, the record it touches, who the agent acts for, the budget, and what comes back. Your agent is allowed to call your billing API. AgentFort decides whether this $4,500 refund should happen. Our AI agent security guide covers the problem in vendor-neutral terms.
The 5 AgentFort controls
| Control | The question it answers | Where to see it |
|---|---|---|
| Identity | Which agents are running, and who answers for each? | AgentFort Discover and AgentFort Identity |
| Access | What may this agent do, and to which records? | AgentFort Access |
| Data protection | What sensitive data is moving through agents? | AgentFort Data Protection |
| Isolation | What happens when an agent is manipulated? | AgentFort Isolation |
| Visibility | What did every agent do, and why was it allowed? | Receipts, shown in the Isolation demo |
Each platform page has the recorded demo, its full transcript, and what the control does not cover.
How AgentFort is deployed
Enforcement runs on the customer's side, next to their services. The console and control plane run in the customer's cloud or network. AgentFort uses a bring-your-own-cloud deployment model; hosted SaaS is not yet available. AgentFort works alongside the API gateway and identity provider a company already runs. Discovery starts with read-only connectors to systems the company already uses, so a first inventory doesn't require rerouting any traffic. Security at AgentFort explains what each control handles and keeps.
Where we are
Explore the five controls through our recorded product demonstrations. AgentFort is recruiting design partners for guided evaluations in their own cloud, using their selected workflows and integrations. AgentFort's own technology is patent pending.
We are recruiting enterprise design partners: companies putting agents onto systems that move money or hold regulated data, who want AgentFort in their own environment and a direct say in what ships next.
The founder
AgentFort was founded by Raghuveer K. He was previously Lead, Security Automation at Amazon, where he spent 12 years in security automation, application security, identity and payments work. He created Design Inspector and Deep Threat Modeler, internal Amazon tools. He holds a master's degree in Artificial Intelligence and is a named inventor on 5 issued US patents in security automation and design analysis:
- US 12,174,963: Automated selection of secure design patterns
- US 12,019,742: Automated threat modeling using application relationships
- US 11,531,763: Automated code generation using analysis of design diagrams
- US 11,507,655: Automatic and predictive source code generation
- US 10,860,295: Automated detection of ambiguities in software design diagrams
You can find him on LinkedIn.
How to work with us
- Request a demo at /demo for a working session built around your agents and your environment.
- Apply to the design partner program at /design-partners to deploy AgentFort with a direct line to the team building it.
- Read the investor page at /investors if you are looking at the company.
- Watch the product in the one-minute film and the 5 recorded demos on our Out of Bounds YouTube channel.
Contact
- General and sales: [email protected]
- Design partners: [email protected]
- Security reports: [email protected], as described on Security at AgentFort
Frequently asked questions
Does AgentFort replace my API gateway or identity provider?
No. AgentFort works with both. It can sit behind any API gateway as the last hop before a service, inside an Envoy-based gateway through ext_authz, or beside the gateway for discovery and identity only. For identity, it uses an agent's existing cloud identity or identity provider, and issues its own SPIFFE identity only when an agent has none.
What open standards is AgentFort built on?
AgentFort is built on open standards: SPIFFE and SPIRE, Envoy, Cedar, OPA, OIDC single sign-on and PostgreSQL. In practice, an agent with no identity can get a SPIFFE identity once its owner approves, teams already running SPIRE keep theirs, and plain-language access rules compile into Cedar policies that owners can read and test. Your platform and security teams can inspect and reason about all of these.
Who is AgentFort for?
Security, platform and identity teams at enterprises putting AI agents to work on real systems: agents that read customer records, call internal APIs, or touch money and regulated data. It is built for the people who have to sign off before an agent reaches production, and who need evidence of what each agent was allowed, denied or asked to review.