IdeGYM

An open-source orchestrator for scalable, disposable development environments - built for training reinforcement learning models and running AI agents.

Already used by JetBrains Research

Try it now!

The problem

Training LLMs with reinforcement learning (RL) on coding tasks requires running tens or hundreds of thousands of generation-reward cycles. Each cycle needs a clean, isolated environment with the right source code and tools.

Scale

A single machine can't run hundreds of parallel environments that reset between iterations in seconds. You need a distributed system built for that from the ground up.

Isolation

Agents must not interfere with each other. Every environment needs its own filesystem, process tree, and network to fully reset between episodes.

Latency

Spinning up a new container for every cycle is too slow at scale. Environments need to restart or reset in under a second to keep training throughput high.

What IdeGYM does

IdeGYM is a Kubernetes-based orchestration framework that manages the full lifecycle of development environments, from image build to teardown.

Environments

Spins up isolated environments on demand and tears them down when finished — no manual cleanup.

Any project, any image

Loads projects from a Git URL, archive, or mounted volume. Builds custom Docker images via a plugin API.

Request forwarding

Proxies requests from your training loop directly to running pods and forwards responses back, so you can compute rewards offline or replay episodes.

Features

WIP features

IDE integration

JetBrains IDEs run headless inside pods and route the IDE's built-in MCP server through the IdeGYM server. This means any MCP-compatible agent gets access to the full IDE toolchain — inspections, refactoring, code intelligence — through the same interface it already uses to call tools and compute rewards.

MCP support

MCP has become the de facto standard for connecting AI agents to external services. IdeGYM exposes its core APIs — tools, rewards, filesystem, and server lifecycle — as MCP endpoints, so any MCP-compatible agent or framework can use IdeGYM environments without writing custom client code.