# AgentFort > AgentFort is security for AI agents, built by Sec-Astra. It finds every AI agent in an organization, gives each one an identity and an accountable owner, checks every tool call (operation, arguments, record, who the agent acts for, budget) against policy, stops sensitive data leaving through agents, isolates code-executing agents in sandboxes, and keeps a tamper-evident receipt for every request. Enforcement runs in the customer's own environment. AgentFort is built by Sec-Astra, an independent company, and is not affiliated with Salesforce Agentforce. Product status (October 2026): all five controls are built and demonstrated end to end in a lab; AgentFort is recruiting enterprise design partners for bring-your-own-cloud evaluations. The console, control plane and enforcement run in the customer's cloud or network. Hosted SaaS is not yet available. Contact: hello@agentfort.net. ## Agent security guides - [AI agent security: a practical guide](https://agentfort.net/agent-security/index.md): AI agent security for architects and CISOs: the attack surface, six controls, where to enforce them, a worked refund example and a 30/60/90-day rollout plan. - [Agent gateway: the four kinds, and what each one can decide](https://agentfort.net/agent-security/agent-gateway.md): An agent gateway can mean an AI gateway, an MCP gateway, an A2A gateway or action-level authorization. Compare what each sees, decides and misses. - [AI agent audit trail: recording what each agent did, for whom, and why it was allowed](https://agentfort.net/agent-security/audit.md): How to build an AI agent audit trail: what to record for each action, what to leave out, hash-chained tamper evidence, framework mapping and OCSF for your SIEM. - [AI agent authorization: deciding what each agent call may do](https://agentfort.net/agent-security/authorization.md): AI agent authorization explained: why roles and scopes fall short, the inputs to each tool call decision, policy models, delegation, MCP and a checklist. - [Coding agent security: how to secure AI coding agents on laptops and in CI](https://agentfort.net/agent-security/coding-agents.md): Coding agent security for Claude Code, Codex, Copilot and Cursor: the main risks, each tool's built-in sandbox and approval controls, and a checklist. - [AI agent data loss prevention: keeping data where it belongs](https://agentfort.net/agent-security/data-protection.md): AI agent data loss prevention for all four data flows: decode before inspecting, choose detectors, mask or hold data, quarantine poisoned documents, log safely. - [Shadow AI agents: how to discover and inventory the agents at work in your organization](https://agentfort.net/agent-security/discovery.md): A practitioner's guide to shadow AI agents: the evidence that reveals them, how to reconcile declared and observed agents, and a two-week agent inventory plan. - [Agent security glossary: AI agent security terms, defined](https://agentfort.net/agent-security/glossary.md): Agent security glossary: plain definitions of 55 AI agent security terms, from MCP, A2A and prompt injection to SPIFFE, Cedar, sandboxes and audit receipts. - [AI agent guardrails: what they catch, where they fail, and what goes behind them](https://agentfort.net/agent-security/guardrails.md): AI agent guardrails explained: what input and output classifiers catch, where they fail, how to test them, and the deterministic controls to put behind them. - [Human in the loop AI agent approval: when to ask and how to bind it](https://agentfort.net/agent-security/human-approval.md): Human in the loop AI agent approval done well: risk tiers, binding approval to the exact call, single use and expiry, separation of duties and approval fatigue. - [AI agent identity: a name each agent can prove, an owner, and who it acts for](https://agentfort.net/agent-security/identity.md): AI agent identity explained: why shared keys and borrowed tokens fail, what each identity must carry, and how to issue one with cloud IAM, an IdP or SPIFFE. - [MCP security: securing the tools your AI agents call](https://agentfort.net/agent-security/mcp-security.md): MCP security in practice: the threat model, what the MCP authorization spec requires, per-call checks, tool pinning, gateways, local servers and a checklist. - [OWASP agentic top 10: controls and tests for ASI01 to ASI10](https://agentfort.net/agent-security/owasp-agentic-top-10.md): The OWASP agentic top 10 (2026) made practical: for ASI01 to ASI10, an example, preventive and detective controls, how to test, and where to enforce each one. - [AI agent sandbox: how to isolate agents that run code](https://agentfort.net/agent-security/sandboxing.md): How to build an AI agent sandbox: assume-breach design, containers vs gVisor vs microVMs vs kernel primitives, egress, credentials, limits and a test plan. - [AI agent spending limits: how to cap refunds, credits and payments](https://agentfort.net/agent-security/spending-limits.md): AI agent spending limits explained: per-call, per-record and per-period caps, currency and unit traps, what happens at the limit, and a worked refund example. - [AI agent security tools: the categories, and what each one covers](https://agentfort.net/agent-security/tools.md): A neutral map of AI agent security tools as of October 2026: 10 categories, what each covers and misses, example vendors, a coverage matrix and 3 stacks. ## Product - [AgentFort Access: least privilege access for AI agents, decided call by call](https://agentfort.net/platform/access.md): Least privilege access for AI agents with AgentFort Access: every REST and MCP call checked for identity, operation, arguments, record, acting-for and budget. - [AgentFort Data Protection: data protection for AI agents, in both directions](https://agentfort.net/platform/data-protection.md): Data protection for AI agents with AgentFort: decode, inspect and mask what agents read and send, keep personal data from models, hold hidden instructions. - [AgentFort Discover: an AI agent discovery tool built on evidence](https://agentfort.net/platform/discover.md): AgentFort Discover is an AI agent discovery tool that finds every agent, including shadow AI, from platforms, model calls, identities, code and runtime. - [AgentFort Identity: AI agent identity management that starts from what you have](https://agentfort.net/platform/identity.md): AI agent identity management with AgentFort: use an agent's cloud identity, plug into your identity provider, or issue a SPIFFE identity the owner approves. - [AgentFort Isolation: an AI agent sandbox that assumes the code is hostile](https://agentfort.net/platform/isolation.md): AgentFort Isolation is an AI agent sandbox built from Linux kernel features: no network, no credentials, one inspected way out, and a receipt for every request. - [Homepage](https://agentfort.net/): overview of the five controls, demo videos and how AgentFort is deployed. ## Articles - [Find the AI agents at work in your business, including the ones nobody approved](https://agentfort.net/blog/find-ai-agents.md): How to find AI agents in your business, including shadow AI: the five places agents leave evidence, six categories, five risk flags and a sample inventory. - [Keep your API gateway for AI agents: where decisions about agent actions belong](https://agentfort.net/blog/keep-your-gateway.md): Your API gateway for AI agents routes, authenticates and rate-limits. Learn where decisions on each agent action belong and three ways to add them. - [How to stop an AI agent issuing a $4,500 refund](https://agentfort.net/blog/stop-ai-agent-refund.md): How to stop AI agent refunds that break your rules: limits per invoice and week, record checks, Finance approval and a receipt for every decision. ## Company - [About AgentFort](https://agentfort.net/about.md): AgentFort is security for AI agents, built by Sec-Astra. Who we are, the 5 controls, how AgentFort is deployed, our founder, and how to work with us. - [AgentFort security: how the product runs and handles your data](https://agentfort.net/security.md): AgentFort security for reviewers: where AgentFort runs, what data each control handles and keeps, how credentials are held, and how to report a vulnerability. - [Request a demo](https://agentfort.net/demo): see AgentFort on your own agents. - [Design partner program](https://agentfort.net/design-partners): deploy AgentFort in your environment with the founding team. ## Optional - [Full text of all guides](https://agentfort.net/llms-full.txt): every page above in one Markdown file.