Documentation

Build governed MCP workflows with z-gateway.

Start with setup, then go deeper on MCP clients, policies, rollout modes, GitHub and Postgres connector security, runtime sessions, and approval patterns.

Start setup MCP setup

Getting started

Create a workspace, connect an agent, and configure MCP.

Getting StartedCreate a workspace, connect an agent, configure MCP, and route GitHub or Postgres actions through z-gateway.MCP SetupConfigure Codex, Claude, or a generic MCP client.

Core concepts

Understand the policy and rollout model.

PoliciesDefine runtime allow and deny rules for agent tool calls.Monitor ModeObserve policy decisions before enforcing blocks.Enforce ModeBlock denied actions before execution.

Connectors

Connect external systems through governed runtime access.

GitHub ConnectorUse GitHub App permissions and runtime policy enforcement.Database ConnectorsConnect Postgres targets for schema inspection, read queries, write proposals, safe writes, risk assessment, and approvals.Runtime SessionsReview GitHub and database actions in one execution timeline.

Security

Review the security model and positioning.

SecurityShort-lived tokens, policy enforcement, audit logs, and redaction.Prompt Guardrails vs Runtime GovernanceWhy tool-call security needs runtime controls.

Use cases

Apply z-gateway to common AI agent workflows.

AI Coding AgentsGovern coding agents across repositories, database inspection, write proposals, and approvals.GitHub MCP SecuritySecure GitHub access for MCP-connected agents.Postgres Agent SafetyControl database-aware agents with read governance, write risk assessment, and approvals.

Recommended path

New teams should create a workspace, register an agent, configure MCP, connect GitHub and/or Postgres, start in monitor mode, review runtime sessions, and enable enforce mode when policy decisions match expected agent behavior.