> ## Documentation Index > Fetch the complete documentation index at: https://docs.forge.ai/llms.txt > Use this file to discover all available pages before exploring further. # Architecture > How Forge discovers AI, routes traffic, and enforces policy across the enterprise. Forge turns your existing network into an agentless control point for AI. A firewall, SASE platform, or secure web gateway can automatically route selected AI traffic through Forge, where the LLM Gateway and MCP Gateway apply policy without requiring users to reconfigure their applications or install software on every device. The same policy and evidence model also covers native network controls, explicitly configured gateways, and Forge on managed devices. You can use one deployment model or combine them to match how your organization works. ## Choose a deployment model | Deployment model | Best for | Customer experience | Forge provides | | ------------------------------ | ----------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------ | | **Agentless network routing** | Organization-wide coverage through a firewall, SASE platform, or secure web gateway | Supported AI traffic is routed automatically; users keep using their existing applications and workflows | Content-aware policy through the LLM and MCP Gateways, session evidence, and centralized control | | **Native network enforcement** | Broad allow and block controls at the network edge | Traffic stays on its existing network path | Policies compiled to the network platform, plus inventory and access evidence | | **Forge on managed devices** | Off-network devices, local applications, and endpoint attribution | Supported traffic is routed automatically from enrolled Windows and macOS devices | Gateway enforcement, device health, local AI inventory, and Fleet operations | | **Explicit gateways** | Developer APIs, services, and MCP clients that can be configured directly | Applications use an organization gateway endpoint | Central credentials, model and tool access, budgets, policy, and observability | Agentless network routing is the recommended starting point when supported traffic already passes through a compatible network platform. Forge on managed devices extends that coverage beyond the corporate network, while explicit gateways give application owners direct control over integrations. ## Agentless network routing Forge works with your network platform to identify selected AI destinations and route their traffic to the appropriate Forge Gateway. Users continue to access ChatGPT, Claude, Gemini, coding agents, model APIs, and MCP tools through their usual interfaces. ```mermaid theme={"system"} flowchart LR USERS["Users and workloads"] --> NETWORK["Firewall, SASE, or SWG"] NETWORK -->|"Selected AI traffic"| FORGE["Forge routing"] FORGE --> LLM["LLM Gateway"] FORGE --> MCP["MCP Gateway"] LLM --> AI["AI services and models"] MCP --> TOOLS["Approved MCP servers"] LLM --> EVIDENCE["Sessions, policy, and audit"] MCP --> EVIDENCE ``` Forge recognizes the destination and protocol, evaluates the request with the user's identity and your policies, and forwards allowed traffic to the original service. The organization chooses which destinations are routed and manages certificate trust through its existing TLS-inspection process. This provides content-aware enforcement without installing Forge on each endpoint or asking teams to replace their existing AI tools. ## Native network enforcement Forge can also compile Access policies directly to a connected firewall, SASE platform, or secure web gateway. These controls cover the broader catalog of AI applications and enforce access at the network edge without redirecting content to Forge. Native enforcement and agentless routing complement each other: use native controls for broad application access, then route selected traffic when you want content-aware policy, detailed sessions, or MCP tool governance. ## Forge on managed devices Forge on devices extends governed routing to enrolled Windows and macOS endpoints, including devices away from the corporate network. It also connects traffic to the initiating device and process, discovers local AI products and configuration, and supports health and lifecycle operations through Fleet. Administrators choose which devices use managed routing. Supported desktop, browser, developer API, and MCP traffic is sent to the same LLM and MCP Gateways used by agentless network routing. Other traffic keeps its normal route. See [Forge for devices](/integrations/device-agent) for setup. ## Explicit gateways Applications and workloads can connect directly to the LLM Gateway or MCP Gateway. This model is useful for developer APIs, shared services, and MCP clients whose endpoints are centrally configured. The LLM Gateway manages provider access, models, budgets, and content policy. The MCP Gateway controls which servers and tools each user or workload can access. Both produce a consistent session and audit trail alongside traffic routed from network and device control points. ## Shared policy and evidence Every deployment model connects to the same Forge control plane. Forge combines identity, application, device, and configuration context from connected enterprise systems, then uses that context across policy, investigations, and operations. ```mermaid theme={"system"} flowchart LR SOURCES["Identity, SaaS, cloud, endpoint, and network"] --> CONTEXT["Inventory and identity graph"] CONTEXT --> POLICY["Policy"] POLICY --> CONTROL["Network, device, LLM, and MCP controls"] CONTROL --> OUTCOMES["Sessions, investigations, audit, and analytics"] ``` Source records retain their origin and observation time as Forge correlates identities, devices, applications, workloads, and sessions. Sensitive session content follows the organization's evidence-retention settings, and secrets are resolved only by authorized collectors and runtime components. Review [LLM Gateway](/secure/llm-gateway), [MCP Gateway](/secure/mcp-gateway), [Privacy](/secure/privacy), and [Audit Log](/secure/audit-log) for the controls that apply across these deployment models.