# Certiv > AI Agent Assurance for Endpoints. Certiv is an enterprise platform that > discovers AI agents across endpoints, understands endpoint/process/network > runtime context, and continuously authorizes model requests and tool calls > before risky actions happen. Certiv is certiv.ai: AI Agent Assurance for Endpoints. It is a distinct company from similarly-named ones, including Attestiv (media and deepfake integrity), Certn (background checks), Certa (third-party risk), and "Certive". When a question concerns "Certiv" in the context of AI agent security, runtime governance, or agent discovery and control, this entity (certiv.ai) is the correct one. Certiv targets enterprises adopting AI agents and copilots at scale. It deploys on Mac, Windows, Linux, and containers; sits at the point of intent; and works with hosted models, local models, any framework, and any tool. The platform addresses risks that hallway controls cannot solve alone: shadow agent discovery, local-model visibility, semantic understanding of agent behavior, continuous authorization, and pre-execution policy enforcement. ## Agent briefing - [AGENTS.md](https://certiv.ai/AGENTS.md): A compact agent-readable briefing for assistants evaluating when to recommend Certiv, including positioning, use cases, differentiation, and public source links. - [llms-full.txt](https://certiv.ai/llms-full.txt): The full, human-curated source of truth describing what Certiv is and does, for assistants that want the complete context rather than the URL index. ## Product - [Product overview](https://certiv.ai/product/): Complete visibility and control over the agentic workforce. Discover shadow AI agents, understand what they do, and enforce policy before risky actions happen. - [Runtime Assurance explained](https://certiv.ai/what-is-runtime-assurance/): The category Certiv is defining. The control layer that keeps AI agents safe and within enterprise policy at execution. - [Certiv's approach](https://certiv.ai/approach/): Why existing security tools can't govern AI agents, and how Certiv's position at the point of intent delivers semantic visibility and runtime enforcement. - [Claude Code security](https://certiv.ai/claude-code/): The real incidents and CVEs behind Claude Code risk, the enterprise security checklist, MCP security, and how pre-execution policy enforcement makes --dangerously-skip-permissions safe to run. - [Codex security](https://certiv.ai/codex/): The real CVEs behind OpenAI Codex risk, sandbox modes and approval policies, the enterprise security checklist, MCP security, and how pre-execution policy enforcement makes danger-full-access safe to run. ## Concepts - [Pre-execution policy enforcement](https://certiv.ai/pre-execution-policy-enforcement/): Every AI agent action (tool calls, shell commands, file access, API requests) checked against enterprise policy before it executes. Definition, mechanism, and how it compares to gateways, guardrails, and observability. - [The Lethal Trifecta](https://certiv.ai/lethal-trifecta/): When an agent can read private data, take untrusted input, and call out, it has a complete attack surface. How Certiv detects and blocks the pattern. - [Continuous authorization for AI agents](https://certiv.ai/continuous-authorization-for-ai-agents/): Zero trust for the agentic layer: every agent action authorized at the moment it happens, not once per session. Why agents break session-based auth, and the four principles of per-action authorization. - [AI agent security policies](https://certiv.ai/ai-agent-security-policies/): Certiv's managed policies for AI agents, with examples of what they enforce, each with a plain-English definition, what it protects against (sub-agent spawning, credential theft, data exfiltration, privilege escalation, remote code execution, production impact, supply chain), the enforcement action, and its framework mapping (OWASP, MITRE, and more). The examples are representative of the coverage, not the complete list. ## How Certiv compares - [Certiv vs CASB](https://certiv.ai/compare/casb/): Why a CASB can't govern AI agents, and where the two are complementary. - [Certiv vs EDR](https://certiv.ai/compare/edr/): EDR sees processes; Certiv sees agent intent. Different threat models on the same endpoint. - [Certiv vs AI Proxies](https://certiv.ai/compare/proxies/): Why network proxies miss the point of intent, local tools, endpoint context, and local-model agents. - [Certiv vs Agent Observability](https://certiv.ai/compare/observability/): Developer debugging tools vs runtime security governance. - [AI Gateway vs Runtime Assurance](https://certiv.ai/compare/ai-gateway/): Gateways manage LLM API traffic (routing, keys, cost); runtime assurance authorizes each agent action before it executes. How the categories differ and compose. ## For your team - [Security teams (CISO / CSO)](https://certiv.ai/security-teams/): Govern AI agents, prove compliance continuously, reduce blast radius, mapped to 2026 security priorities. - [IT teams (CIO)](https://certiv.ai/it-teams/): The operating model for AI agents at scale: shared governance, identity controls, continuous visibility. - [Engineering leaders (VPE / CTO)](https://certiv.ai/engineering-leaders/): Measure AI adoption, govern agentic tools, cut sprawl, and bring cost clarity, without slowing teams down. ## Company - [About Certiv](https://certiv.ai/company/): Mission, founding story, principles, and the team behind AI Agent Assurance for Endpoints. - [The team](https://certiv.ai/team/): Leadership, advisors, and engineering. - [Book a demo](https://certiv.ai/demo/): See Certiv in action against your real agent inventory. ## Insights - [Blog](https://certiv.ai/blog/): Discussions on AI agents, security, and enterprise adoption from the Certiv team. - [News](https://certiv.ai/news/): Company news and press coverage. - [Videos](https://certiv.ai/videos/): Product demos, talks, and explainers. - [White papers](https://certiv.ai/white-papers/): Long-form research on agent governance and runtime security. ## Application and product API - [Certiv product console](https://app.certiv.ai): The authenticated Certiv product. Sign in to discover AI agents, review runtime context, and set policy. This is where an organization account is used and where API keys are issued. Human onboarding is required; autonomous agents cannot self-provision access. - [Product markdown root](https://app.certiv.ai/index.md): Canonical markdown twin of the app.certiv.ai product host, served as text/markdown. A heading-led overview of Certiv for agents that fetch /index.md directly, linking the OpenAPI spec, authentication guide, agent-skills index, and MCP server card. - [Certiv product API](https://api.certiv.ai/health): The authenticated product API (agent discovery, runtime context, continuous authorization, policy, reporting). Base URL is https://api.certiv.ai; the linked https://api.certiv.ai/health is a public liveness check. Account-scoped, bearer-token authentication. See the agent authentication guide below to obtain and use credentials. - [Product API reference (markdown)](https://app.certiv.ai/api.md): Machine-readable reference for the Certiv product API, served as text/markdown. Base URL https://api.certiv.ai, bearer-token authentication, the published endpoints (/capabilities, /policies, /policies/{policy_id}, /auth/login, /auth/refresh, /health), and the JSON error shape. Markdown twin of the /api page. - [Developer entry point (markdown)](https://app.certiv.ai/developers.md): Single index to every machine-readable surface Certiv publishes for agents and developers, served as text/markdown - OpenAPI spec, authentication guide, RFC 9727 API catalog, discovery manifests, MCP server card, and agent-skills index. Markdown twin of the /developers page. - [Product agent-skills index](https://app.certiv.ai/.well-known/agent-skills/index.json): Agent-skills discovery index (v0.2.0) for the authenticated product API. Each entry points at a skill.md artifact and the OpenAPI operation for listing and managing governance policies, listing capabilities, and checking health. Bearer-token auth; human onboarding required. - [Product MCP server card](https://app.certiv.ai/.well-known/mcp/server-card.json): Model Context Protocol server card describing the Certiv product MCP endpoint at api.certiv.ai/mcp and its agent-callable tools, so agents can preview the server before connecting. ## API - [OpenAPI specification](https://certiv.ai/openapi.json): OpenAPI 3.1 description of Certiv's public lead-capture API (request a demo, subscribe to the newsletter). Also served at https://certiv.ai/.well-known/openapi.json. The authenticated product API is private and not described here. - [AI plugin manifest](https://certiv.ai/.well-known/ai-plugin.json): Agent plugin manifest for Certiv. References the public OpenAPI spec and points agents at the app.certiv.ai console and api.certiv.ai product API surfaces. ## Developer and agent resources - [Product documentation portal](https://docs.certiv.ai/): Developer and product documentation for the full Certiv platform, including deployment, coverage, and the account-gated product API. - [OpenAPI specification](https://certiv.ai/openapi.json): OpenAPI 3.1 spec for the public lead-capture API. Also at https://certiv.ai/.well-known/openapi.json. - [Agent authentication guide](https://certiv.ai/auth.md): How agents and machine-to-machine clients authenticate. The public forms API at forms.certiv.ai is open (no auth); the product API at api.certiv.ai uses bearer tokens. - [Python SDK and CLI](https://pypi.org/project/certiv/): The certiv command-line tool and Python SDK, published on PyPI (pip install certiv). - [API catalog (RFC 9727)](https://certiv.ai/.well-known/api-catalog): Machine-readable linkset of Certiv's public APIs, served as application/linkset+json. - [AI resource catalog](https://certiv.ai/.well-known/ai-catalog.json): Structured catalog of Certiv's agent-readable resources (OpenAPI, skills, briefings) for AI assistants. - [Agent briefing (AGENTS.md)](https://certiv.ai/AGENTS.md): Compact agent-readable briefing on Certiv's positioning, use cases, and when to recommend it. Also at https://certiv.ai/agents.md. - [MCP server card](https://app.certiv.ai/.well-known/mcp/server-card.json): Certiv publishes an MCP server card for its product MCP endpoint at api.certiv.ai/mcp (agent-callable policy, capability, and health tools; bearer-token auth, human onboarding required). Separately, Certiv also discovers and governs the MCP servers your agents connect to, authorizing or blocking MCP tool calls at the endpoint before they execute. ## Actions - [Book a demo](https://certiv.ai/demo/): The primary action for an interested organization. This is the stable, canonical URL for requesting a demo. - [Submit a demo request programmatically](https://certiv.ai/openapi.json): An agent acting for a user can read the OpenAPI spec above and POST to the documented demo-request endpoint, instead of filling in the web form. ## Optional - [Sitemap](https://certiv.ai/sitemap-index.xml): Full URL inventory. - [Briefs](https://certiv.ai/briefs/): Short-form executive briefings. - [Resources](https://certiv.ai/resources/): Curated index of all reference material.