Unified AI System 0.4.0: a governed, terminal-first MCP gateway with local prompt enhancement

[happy520ai]

Pre-submission Checklist

• This post follows the MCP community guidelines

What would you like to share?

Hi, I maintain Unified AI System, an Apache-2.0, terminal-first AI and MCP gateway for models, agents, knowledge, and tools.

Version 0.4.0 is active in the official registry as io.github.happy520ai/unified-ai-system. The public evaluation path is deliberately credential-free: the MCP server starts an isolated managed gateway, exposes nine tools, uses a deterministic local fake provider for chat, and removes the managed process when the session ends.

The part I would most value feedback on is the intent-to-execution boundary. gateway_prompt_enhance turns a rough natural-language request into a structured prompt locally, preserves the original request, and makes no provider call. Applying the enhanced prompt to chat remains explicit; ordinary chat behavior is unchanged.

What works today:

• Nine stdio tools covering gateway health/readiness, provider-free prompt enhancement, safe fake-provider chat, knowledge readiness, workflows, and workforce status.
• Real-provider execution is disabled throughout the default MCP path and fails closed if a gateway may call a real provider.
• An anonymously pullable multi-architecture container; no clone, account, or API key is needed for the first handshake.
• Codex, Cursor, and Cline setup guides plus a separately installable Codex plugin and agent skill.
• Public-clone verification covers initialization, tool discovery, calls to all documented tools, fake execution, and managed-process cleanup.

Try the MCP server with Codex:

codex mcp add unified-ai-system -- docker run --rm -i ghcr.io/happy520ai/unified-ai-system/mcp-server:0.4.0

Then ask Codex:

Use gateway_prompt_enhance to turn "build a small API for my team" into a coding prompt. Show the preserved original request, detected language, enhanced prompt, and proof that no provider was called.

Current boundaries are explicit: this is a self-hosted public preview, not a hosted API, production certification, L5 autonomy, or AGI claim. The default path proves governed local behavior before anyone supplies provider credentials.

I would appreciate technical feedback on two questions:

1. Should deterministic prompt enhancement remain a dedicated tool, or is it more natural as an optional field on a chat tool?
2. Which evidence would you expect a gateway to expose before allowing a client to opt into real-provider execution?

If this direction is useful, feedback, testing, or a Star would help other builders discover it.

Relevant Links

• Repository: https://github.com/happy520ai/unified-ai-system
• Official Registry entry: https://registry.modelcontextprotocol.io/v0.1/servers/io.github.happy520ai%2Funified-ai-system/versions/0.4.0
• Release: https://github.com/happy520ai/unified-ai-system/releases/tag/v0.4.0
• 60-second Codex MCP quickstart: https://github.com/happy520ai/unified-ai-system/blob/master/docs/codex-mcp-quickstart.md
• Prompt-enhancement behavior: https://github.com/happy520ai/unified-ai-system/blob/master/docs/prompt-enhancement.md
• Public verification workflow: https://github.com/happy520ai/unified-ai-system/actions/workflows/ci.yml