Phelix - Zero-downtime deploys for Go & Rust, with MCP for AI agents
by•
Phelix is a single-binary deployment tool for Go and Rust apps that runs on your own servers. Build,version, and deploy with zero downtime, health-check-gated traffic switching, encrypted environment snapshots, and instant rollback. Use blue-green, rolling, or canary rollouts with or without Docker. Built-in MCP support lets coding agents inspect services and prepare deployment plans through Phelix's existing authorization and approval pipeline.An optional dashboard adds multi-server monitoring.

Replies
Hey Product Hunt! I'm Mohammad, the solo developer behind Phelix.
I built Phelix because deploying services on my own servers kept turning into the same routine: build the app, replace the running process, check whether it came back, and hope I had a rollback path if something went wrong.
Kubernetes felt like more infrastructure than I needed for a small fleet of services. But a collection of shell scripts wasn't the deployment workflow I wanted either.
So I built Phelix as a single binary for Go and Rust applications. It versions builds on disk, keeps the current version running until the new one passes its health checks, snapshots encrypted environment variables per version, and makes rollback a deliberate, repeatable operation.
Since v1.0.0, I've continued improving the deployment workflow. Phelix is now at v1.1.2, and one of the biggest additions is built-in MCP support for coding agents.
Why MCP?
AI agents are getting better at writing code, but shipping that code safely is a different problem.
Phelix gives compatible agents such as Claude Code and Codex a controlled way to inspect application context, create deployment or rollback plans, and apply plans through Phelix's existing execution pipeline.
Plans are immutable and content-addressed. Execution remains behind the authorization boundary, and deployments can require explicit human approval. MCP doesn't introduce a separate deployment engine or bypass the existing safeguards.
The MCP server runs locally over stdio, without opening a network listener. The goal is simple: let agents help with deployment work without making production changes an unreviewed side effect of a prompt.
Phelix is still a solo-built project, and I'm building it around real backend workloads rather than claiming it has been battle-tested everywhere.
I'd love your feedback on two things:
How do you currently deploy Go or Rust services on your own servers?
Would you let a coding agent prepare a deployment plan if you could inspect and approve the exact plan before execution?
Thanks for checking out Phelix. I'm happy to answer technical questions and hear what you'd change.