Lineation

About Lineation

Lineation is a security control plane for AI agents, positioned for the gap between traditional perimeter security and agents that read documents, call APIs, and move money on their own. It organizes agent security into three layers you can adopt in order. Observability gives real-time dashboards across every provider, reasoning and decision logs, auto-discovery of shadow AI nobody registered, and a replay mode that functions as a black box recorder for agent runs. Governance adds policy-as-code evaluated at runtime, RBAC and approval gates, zero-trust non-human agent identity so every agent has a scoped, owned, continuously authorized credential instead of a shared API key, and templates for SOC 2, HIPAA, and the EU AI Act. Defense covers prompt-injection scrubbing, PII and secret redaction at output time, sandboxed execution, and a kill switch for compromised agents. Two gateways make it enforceable: an LLM gateway that normalizes and policy-checks requests before forwarding to OpenAI, Claude, Gemini, or Copilot, and an MCP gateway applying the same identity, policy, and audit to MCP tool and server traffic so tool calls are mediated like model calls. The architecture is central governance with distributed enforcement — decide in one place, prove everywhere. Billing is per protected action, defined as any brokered step: evaluating a prompt, calling a tool or API, hitting a database, or delivering a final response. Internal chain-of-thought loops do not consume quota. On-prem and air-gapped deployments are offered.

Key Features

Lineation Pros & Cons

Who Is Lineation Best For?

Tags