Show HN: Run LLMs in Docker for Any Language without Prebuilding Containers

Why It Matters

By containerising LLM‑powered coding assistants, agent‑en‑place eliminates version drift and security risks, enabling teams to adopt AI tools without compromising reproducibility or compliance.

Key Takeaways

• Auto-detects language versions from common config files
• Generates Docker images with exact tool versions on demand
• Supports Codex, Opencode, GitHub Copilot providers
• Caches images to speed up repeated runs
• Offers debug, rebuild, and Dockerfile preview flags

Pulse Analysis

The rapid rise of large‑language‑model (LLM) coding assistants has created a paradox for developers: while these tools promise to accelerate code generation, they also introduce dependency hell. Different projects rely on distinct Node, Python, Ruby or Go versions, and mismatched environments can cause subtle bugs or outright failures. Traditional virtual environments or manual Dockerfiles demand constant upkeep, especially when teams switch between providers like OpenAI’s Codex, Opencode‑AI, or GitHub Copilot. Agent‑en‑place addresses this friction by automating the discovery of version specifications directly from existing project files, ensuring that the container mirrors the developer’s intended stack.

Under the hood, the CLI scans for .tool‑versions, mise.toml and language‑specific version files, then synthesises a minimal Debian‑12‑slim Dockerfile that installs the mise runtime manager and the exact tool versions required. The image is built on‑the‑fly—or retrieved from a local cache—using a deterministic naming scheme that reflects the included toolset. When executed, the container mounts the current workspace and the provider’s configuration directory, preserving authentication tokens and settings while running the LLM tool as a non‑root user for added security. Advanced flags let users expose the Docker build log, force a rebuild, or simply output the generated Dockerfile for custom tweaks.

For enterprises, this approach offers tangible operational benefits. Containerising AI assistants isolates them from host environments, reducing attack surface and simplifying compliance audits. Cached images cut down CI/CD pipeline latency, and the ability to switch providers with a single command encourages experimentation without lock‑in. As AI‑driven development becomes mainstream, tools like agent‑en‑place will likely become a standard part of the DevOps toolkit, bridging the gap between cutting‑edge LLM capabilities and the rigorous stability requirements of production software.