FAQ

Common questions about Agent Apps.

1. What's the difference between Agent Apps and MCP servers?

MCP servers give your agent tools — functions it can call. Agent Apps give your agent tools + interfaces + lifecycle + events + storage + app-to-app communication.

MCP (Model Context Protocol) is a protocol for exposing functions to AI clients. It's great for stateless tool providers. But it doesn't handle authentication flows, visual interfaces, persistent storage, real-time events, app-to-app communication, or runtime management.

Agent Apps builds on top of MCP. It uses MCP as its tool-calling layer (via the Gateway MCP bridge), so Agent Apps work in any MCP-compatible client. But behind the tools, there's a full application runtime.

→ See Agent Apps vs MCP Servers for the full comparison.

2. Is Agent Apps production-ready?

Agent Apps is currently at v0.1.0 — an early pre-release of the protocol, SDK, and toolchain. The 8 packages (@agentapps/protocol, @agentapps/sdk, @agentapps/station, @agentapps/cli, @agentapps/host, @agentapps/create, @agentapps/execution-host, @agentapps/testing) are available from the GitHub repo and will be published to npm in a future release.

The 80+ apps in the catalog range from fully functional to stub implementations — each app's maturity varies. Check the app's manifest and tools before relying on it in production.

The Station supports production features: multi-tenancy, Docker and Kubernetes runtime adapters, remote Execution Hosts, health monitoring, backpressure, and restart resilience.

3. Can I use Agent Apps with ChatGPT / Copilot / [client]?

Agent Apps works with any MCP-compatible client. Currently tested and configured for:

• Claude Code ✅
• Claude Desktop ✅
• Cursor ✅
• Windsurf ✅
• VS Code (Copilot Agent Mode) ✅
• Zed ✅

For clients that don't support MCP yet (like ChatGPT's web interface), Agent Apps can't be used directly — the client needs to support MCP or the Agent Apps protocol natively. As more clients add MCP support, they'll work with Agent Apps automatically.

4. Do I need to rewrite my existing MCP servers?

No. Existing MCP servers continue to work as-is. You have three options:

1. Keep using them — they work fine for simple, stateless tools
2. Wrap them — expose an MCP server as an Agent App to get lifecycle, auth, and storage
3. Rewrite them — use the @agentapps/sdk to build a native Agent App with interfaces, events, and app-to-app communication

The SDK is designed to be as simple as writing an MCP server, but with the full application layer included.

5. How do I deploy a Station?

For local development:

agentapps serve --port 4100

For production, the Station supports multiple runtime adapters:

Configuration lives in agentapps.json:

{
  "station": {
    "publicUrl": "https://apps.yourdomain.com",
    "port": 4100
  },
  "apps": [
    { "source": "./apps/google-gmail" },
    { "source": "./apps/slack" }
  ]
}

→ See Managing Agent Apps for the full guide.

6. Is there Python / Go / Rust support?

The Agent Apps SDK is currently TypeScript/Node.js only. The protocol itself is language-agnostic (JSON messages over HTTP), so SDKs in other languages are possible.

The protocol is open and the message schemas are defined with Zod in @agentapps/protocol. If you want to build a SDK in Python, Go, or Rust, the protocol spec is the place to start.

There are no official plans for other language SDKs yet, but community contributions are welcome.

7. Is there a hosted/managed option?

Not yet. Agent Apps is currently self-hosted: you clone the repo, run the CLI, and start a Station on your own machine or server.

A hosted/managed Station is on the roadmap. The architecture is designed for it — the Station already supports multi-tenancy, remote Execution Hosts, and OAuth — but no hosted offering is available today.

8. How is this different from LangChain / CrewAI / [framework]?

LangChain, CrewAI, and similar frameworks are agent orchestration tools — they help you build AI agents that reason, plan, and chain calls together.

Agent Apps is a tool platform — it gives agents the tools to act in the real world. It's not an agent framework; it's what agent frameworks (or standalone agents) use to give their agents capabilities.

You can use Agent Apps alongside any agent framework. The agent (whether built with LangChain, CrewAI, or a raw LLM API) calls Agent Apps tools via MCP. Agent Apps handles the tool execution, authentication, and lifecycle.

9. What's the license?

MIT. The entire Agent Apps project — protocol, SDK, Station, CLI, and all 80+ apps in the catalog — is licensed under the MIT License.

Copyright (c) 2026 Teros.

You're free to use, modify, distribute, and sublicense it. See the LICENSE file for details.

10. How do I contribute?

Agent Apps is open source and contributions are welcome.

1. Fork the repo — github.com/agentapps/agentapps
2. Clone and install:

   git clone https://github.com/agentapps/agentapps.git
   cd agentapps
   ./install.sh

3. Make your changes — follow the development guidelines:
   • Code comments in English
   • Conventional commits (feat:, fix:, perf:, refactor:, chore:)
   • Never push without explicit consent
   • No Co-Authored-By trailer
4. Test your changes:

   pnpm test      # vitest
   pnpm lint      # biome

5. Add a changeset — pnpm changeset with your PR
6. Open a PR — describe what changed and why

Contributing a new app

1. Scaffold a new app: npm create @agentapps
2. Define your manifest (agentapp.json)
3. Implement your tools with @agentapps/sdk
4. Test with @agentapps/testing
5. Validate: npx agent-app validate

→ See Creating an Agent App for the full guide.

Logo policy

Logos are sensitive branding elements. Never generate logos with AI. Source them from official brand resources (press kits, brand guidelines, official websites). If official assets aren't available, leave the placeholder path and flag it as a TODO.