Concepts

The five words you’ll see on every page

• Identity — the name of whoever (or whatever) is making the request. For a human that’s a login; for an AI agent it’s usually a short string like agent-abc123, proven with a token instead of a password.
• Policy — the rule book. A list of “this identity can/can’t do this specific thing.” Wardline checks every request against it.
• Budget — a counter with a limit, like a monthly data cap. Once an identity uses up its allowance, further requests get turned away until the counter resets.
• Audit — the receipt. Every decision Wardline makes, allowed or denied, gets written down with who/what/when — so nothing happens that can’t be looked up later.
• MCP — Model Context Protocol, the common language AI agents use to call external tools. Wardline understands it, but also works with plain HTTP and gRPC — MCP is the main case, not the only one.

With those five in hand, the rest of this section is the mental model behind Wardline: how it’s built, how identity and policy decisions work, which policy backend to pick, and where the audit trail goes. Read this once — every feature page after this assumes it.

• Architecture
• Identity and Policy
• Policy Backends
• Audit Log