AI-Friendly CLI / MCP Interface

[Vadime]

Feature Request: AI-Friendly CLI / MCP Interface for Infrastructure Automation

Is your feature request related to a problem? Please describe.

As AI coding agents and autonomous developer tools become more common (Codex, Claude Code, Hermes, OpenHands, etc.), there is currently no clean or officially supported way for AI systems to interact with QuickStack programmatically.

At the moment, most interactions require:

• manual dashboard usage
• direct SSH access
• custom scripts
• unsafe kubectl/docker access

This makes it difficult to safely automate infrastructure tasks using AI agents.

Examples:

• deploying applications
• restarting services
• reading logs
• creating projects
• provisioning databases
• restoring backups
• managing domains/environment variables
• checking deployment health
• scaling services

Currently the only alternative is often giving AI agents partial root/SSH access to the server, which is not ideal from a security or architecture perspective.

---

Describe the solution you'd like

An official QuickStack CLI and/or MCP (Model Context Protocol) interface designed for AI and automation tools.

For example:

bash quickstack projects list quickstack deploy restart my-app quickstack logs my-app quickstack backup create postgres-prod quickstack env set my-app KEY=value quickstack project create

And optionally:

bash quickstack ai login quickstack mcp start

This would allow AI agents to safely interact with QuickStack through a controlled abstraction layer instead of direct infrastructure access.

Potential capabilities:

• Project management
• Deployments
• Rollbacks
• Backup management
• Log streaming
• Metrics/health checks
• Environment variable management
• Database management
• Domain management
• Service scaling
• Queue/job inspection
• Read-only vs execute permissions
• Audit logs for AI actions

---

Describe alternatives you've considered

Current alternatives include:

• SSH access
• direct kubectl access
• browser automation/playwright
• custom wrappers/scripts around QuickStack internals

These approaches are:

• brittle
• insecure
• difficult to maintain
• hard to permission correctly
• not AI-native

---

Additional context

There is a rapidly growing ecosystem around AI agents operating developer infrastructure:

• OpenAI Codex
• Claude Code
• Hermes
• OpenHands
• Cursor Agents
• MCP-compatible tooling

QuickStack would be an ideal platform for AI-native infrastructure management if it exposed a secure CLI/API/MCP layer.

This could become a major differentiator for QuickStack in the future, especially for self-hosted users running AI-assisted workflows.

[biersoeckli]

Hey @Vadime,

thanks for the detailed feature request. I agree that this could be an interesting feature.

Before building a CLI or MCP interface properly, QuickStack should probably expose a stable REST API first. There is already an issue for that here: #30

Once that API exists, a CLI and later an MCP server could be built on top of it.

If you’re interested in implementing this, feel free to open a PR :)

If not, I can take a look myself, but it will probably take a few weeks before I get to it.

[biersoeckli]

Hey @Vadime

I just released v0.0.13 wich now contains a beta version of the API with OpenAPI spec. From here further changes can be made to fulfill this issue.

If you're interested in contributing, feel free to create a PR.