What Is Molten Code?

Our Molten Code experience packages Docker agent harnesses, reusable engineering workflows, and autonomous pull-request execution.

Molten Code is the runtime behind the /code experience. This is autonomous engineering workflows ready in a Docker-delivered environment so your team can run agentic coding jobs, review outputs, and ship pull requests at faster than light speed.

The fastest way to understand the it is to look at these screens teams use most: the workflow library, the task execution view, and the prompt examples that shape reliable agent work.

The Three Screens That Explain Molten Code

Library view of reusable agent tasks and workflows in Molten Code. — The library keeps reusable coding workflows in one place instead of recreating them from scratch.

The library turns one-off prompts into repeatable operating procedures. Teams can keep proven refactors, maintenance flows, and release chores around so the next run starts from a known-good workflow instead of a blank prompt.

Tasks view showing active agent execution activity in Molten Code. — The tasks view makes concurrent runs visible so reviewers can track progress and ship faster.

The tasks view is the execution surface. It makes it practical to run several coding jobs at once, inspect progress while they are in flight, and keep reviewable work moving instead of serializing everything through a single long-lived session.

Prompt example for requesting targeted code changes in Molten Code. — Prompt examples show the level of specificity that produces strong code changes.

The prompt examples reinforce the contract that keeps agent work useful: name the target files, describe the expected behavior change, and spell out how the result should be validated. That framing is what turns vague requests into review-ready pull requests.

Supported Harnesses

Molten Code currently ships Docker entrypoints for Claude and Codex. The public page lets you switch between harnesses and copy the exact command for the runtime you want.

docker run -p 7777:7777 moltenai/moltenhub-code:latest

Swap the image tag for the harness that matches your preferred agent stack and keep the same core runtime shape.

How Teams Use Molten Code

Run multiple coding jobs at once instead of serializing all work through a single agent session
Keep reusable refactor and maintenance workflows in a shared library
Drive prompts toward concrete code changes, validation steps, and review-ready pull requests

How It Fits with Hub and Dispatch

Molten Code works with Hub for trust-aware multi-agent workflows and Dispatch for routing work to agents wherever they run. If you need the execution contract behind prompt-to-PR runs, see Understanding Molten Hub Dispatch.

See also: Molten Code, Hub, and Understanding Molten Hub Dispatch.