What is Clevername?
Clevername is the agent governance platform that locks every AI agent to its governed mandate — continuous enforcement, tamper-proof audit, zero drift.
Agent Governance Platform
Clevername governs AI agents from day one through every action they take. Every agent passes through a structured security intake and governance review before activation. Once live, its guardrail profile is checked on every tool call — continuously, not once. Drift from the governed mandate is detected and alerted by default, and blocked once enforcement is enabled for the agent.
Bring Your Own Key (BYOK) — every tier
Clevername is fully BYOK on every plan, including Free. Add your own API keys for Anthropic, OpenAI, Google, Azure OpenAI, or any LiteLLM-supported provider, then switch between models freely. Keys are stored in GCP Secret Manager (AES-256-GCM in local dev) — the database holds only a masked hint. You can store multiple keys per provider with labels and bind specific keys to specific agents for cross-agent isolation. There is no managed-LLM tier and no credit system.
Audit-First Architecture
Every API call, tool invocation, and model interaction is logged. Retention is plan-based: 7 days on Free, 30 days on Pro, 365 days on Team, unlimited on Enterprise (with optional SIEM export to Splunk, Datadog, Sentinel, or Elastic). You get full visibility into what your AI tools are doing and who initiated each action. This audit trail is the foundation for compliance and governance features like the Agent Review.
AI Governance
On Team and Enterprise, Clevername adds organizational controls: role-based access, the Agent Review board with multi-reviewer voting, the SOC console, and AI Activity cost attribution. Enterprise additionally adds SSO (OIDC / SAML), SCIM 2.0 provisioning, Shadow AI discovery (with EDR integrations), data labels (Microsoft Purview, Google DLP, AWS Macie), external secret backends, and self-hosted deployment.
Gateway tokens — govern model calls (base URL swap)
This is the first of two integration paths, and the one to reach for when you want to govern an agent's model calls. Every governed agent gets its owncn-*gateway token bound to that agent's org and approved scope. Change your OpenAI-compatible client's base URL to the Clevername gateway (/api/hub/v1) and use the token as the API key — every request is then scanned by CleverGuard, audited, and checked against the agent's guardrail profile and learned baseline. Works with LangChain, CrewAI, the OpenAI / Anthropic / Google SDKs, or any HTTP-capable framework.
The MCP Gateway — govern tool use
This is the second integration path, and the one to reach for when you want to govern an agent's tool use. The MCP Gateway is a single HTTP endpoint that aggregates all your installed MCP servers. It routes tool calls, enforces access policies, and logs everything. Connect your IDE or MCP client once and get all your tools through a single endpoint. It is a separate surface from the base URL swap above — use either or both.
Agent Review
The Agent Review gates agent deployment through structured security review. Every agent fills out a 43-question intake across 8 security domains, reviewers vote, and a governed guardrail profile is checked on every tool call. Drift from the mandate is detected and alerted by default, and blocked once enforcement is enabled.
CleverGuard Scanning
CleverGuard scans every request for prompt injections, PII, secrets, and policy violations. Four tiers of detection — data labels, regex, ML classifier, and LLM review — ensure high recall without sacrificing precision. False positives can be disputed and resolved.