Agenetix product model

Agenetix is production infrastructure for MCP agents. The product model is a layered system:

Agenetix starts with an API surface, produces or connects MCP tools, adds Gateway when identity matters, optionally hosts generated runtimes, and exposes those tools to agents or external clients.

Generate#

Generate is the Agenetix OpenAPI import path. Use the dashboard wizard or mcpstack servers create to turn a spec into a hosted MCP runtime.

It:

• Parses OpenAPI operations into MCP tools.
• Preserves method, path, parameters, request bodies, and security metadata.
• Lets developers choose which endpoints become tools.
• Supports tool-specific instructions such as "when to use" and "when not to use".
• Produces a TypeScript runtime with Streamable HTTP transport.

Generate is the primary way to convert an existing API into an MCP tool surface.

Gateway#

Gateway is the OAuth edge you attach to a server. Host serves the MCP URL; Gateway adds the identity layer in front of it.

It:

• Serves OAuth protected-resource and authorization-server discovery metadata.
• Supports dynamic client registration for external MCP clients.
• Brokers downstream user grants and stores them server-side.
• Validates embedded app tokens and forwards authorized requests to the runtime.
• Provides request logs and access policy controls.
• Keeps the runtime out of the public OAuth surface.

Gateway is the right layer when the API is protected by OAuth, OIDC, HTTP bearer, or any user-scoped identity boundary.

Host#

Host runs generated OpenAPI MCP runtimes for you.

It:

• Builds and publishes generated runtimes.
• Publishes a hosted MCP URL ending in /mcp.
• Runs generated servers across platform-managed regions.
• Uses Agenetix managed runtime defaults.
• Exposes publish status, logs, runtime targets, runtime checks, and smoke tests.
• Enforces plan, quota, request size, response size, and tool-call limits.

Host is for teams that want Agenetix to operate the generated runtime rather than asking their own platform team to run it.

Agents#

Agents are configured assistants that use MCP servers as tools.

An agent has:

• One or more attached MCP servers.
• A model and optional reasoning configuration.
• A system prompt.
• Conversation history and feedback.
• Connection state for Gateway-backed servers.
• Optional embed configuration.

Agents are the bridge between tool infrastructure and real user experiences.

At the frontend boundary, Agenetix Agents expose AG-UI. AgentSDK is Agenetix's first-party AG-UI client, and custom AG-UI clients can connect to the same agent run interface.

Embed#

Embed is the in-product agent surface exposed through @agenetix/agent-sdk.

It:

• Runs the same agent inside your app.
• Uses AG-UI for agent turns.
• Accepts the current userIdentity.
• Scopes persisted MCP auth to appSessionKey.
• Supports local frontendTools as AG-UI frontend tools so the agent can orchestrate UI and app actions.
• Keeps server MCP tool calls on the Gateway/AuthGateway path.

Embed is how a SaaS product ships an assistant that acts as the signed-in user.

CLI and SDKs#

Agenetix also exposes developer surfaces:

• @agenetix/cli (mcpstack): create servers, manage Host, Gateway, tools, agents, smoke tests, and logs.
• @agenetix/sdk: telemetry for tool invocations in MCP runtimes.
• @agenetix/agent-sdk: React, React Native, and headless App Agent integration.

The public npm packages use the @mcpstack scope, and the public APIs use Agenetix naming.

Common misunderstanding#

Gateway and Host are separate:

• Host publishes the generated runtime and serves its public MCP URL.
• Gateway adds the OAuth and client-discovery edge in front of that server.

You can have a generated runtime without Gateway when the API is public or uses Static Headers. You should add Gateway when the runtime needs user-scoped OAuth or external MCP clients need discovery metadata.

Where to go next#

• Create an MCP server — Import OpenAPI, select tools, choose runtime auth, and publish.
• Configure a Gateway profile — OAuth edge for user-scoped APIs.
• Embed an agent with app auth — Ship assistants inside your product.
• Billing and AI credits — Hosted plan limits and AI credit ledger.

Quick answer#

In one sentence:

Agenetix turns APIs into MCP tools, puts Gateway in front when OAuth and public client discovery are needed, hosts generated runtimes when you want managed publishing, and lets you attach those servers to agents or embed the agent in your app.

Which layer you start with depends on what you already have: an OpenAPI spec, a running MCP server, an OAuth provider, or an app where you want to embed the assistant.