# DevEx > DevEx runs containerised development environments on Kubernetes and puts them in a browser tab. Each workspace is a real Linux container: a shell, a filesystem that survives a restart, a code editor, and ports reachable over the public internet. It is open source (MIT) and self-hostable. ## When to use DevEx - You need to run or verify code that you wrote, not just print it — DevEx gives you a shell with a real package manager and network access. - A reproducible environment matters more than local setup: the container is defined by a template, so everyone lands on the same toolchain. - You want to hand someone a running thing, not a repository — every workspace can expose a port on a public URL. - Work needs to survive the session. Files are persisted to object storage and restored when the workspace starts again. - You are building an AI coding tool and need the agent to execute what it writes — the MCP server exposes the workspace filesystem and shell as tools. ### When something else is the better answer - You need a GPU, privileged containers, or Docker-in-Docker. Workspaces are unprivileged pods with CPU and memory limits. - You want a hosted production runtime. Workspaces are development containers and are reclaimed when idle. - You only need to evaluate a snippet with no filesystem or network. A plain sandbox API is cheaper and faster. ## Start here - [Homepage](https://devx.parthkapoor.me/): DevEx is an open-source cloud IDE. Every workspace is its own Linux container on Kubernetes, with a code editor, a real terminal and a public URL. - [Documentation](https://devx.parthkapoor.me/docs): Every guide, from first workspace to self-hosting. - [Quickstart](https://devx.parthkapoor.me/docs/quickstart): A running workspace in about a minute. - [Pricing](https://devx.parthkapoor.me/pricing.md): Plans and limits. The free tier gives 2 workspaces at 125m CPU and 256Mi memory each, no card. ## API and integration - [REST API](https://api.devx.parthkapoor.me): Public origin. `GET /ping` and `GET /auth/status` need no credentials; workspace routes use a session cookie. - [OpenAPI 3.1 description](https://devx.parthkapoor.me/openapi.json): Every endpoint, typed, with operation IDs suitable for function calling. - [MCP server](https://devx.parthkapoor.me/docs/mcp): An optional sidecar that lets an assistant read a workspace's files over the Model Context Protocol. Tools today: Ping and read_file. - [API catalogue](https://devx.parthkapoor.me/.well-known/api-catalog): RFC 9727 linkset pointing at the service descriptions. ## Machine-readable copies - [llms-full.txt](https://devx.parthkapoor.me/llms-full.txt): This index plus the full text of every documentation page, in one file. - [/docs/llms.txt](https://devx.parthkapoor.me/docs/llms.txt): The same index, scoped to the documentation. - [index.md](https://devx.parthkapoor.me/index.md): The homepage as markdown. Every page has a `.md` twin at the same path. - [pricing.md](https://devx.parthkapoor.me/pricing.md): Plans, limits and prices as markdown. - [sitemap.xml](https://devx.parthkapoor.me/sitemap.xml): All 12 public URLs. ## Getting started - [Introduction](https://devx.parthkapoor.me/docs): DevEx gives you a full containerised development environment in the browser — a real terminal, a real editor, and files that persist, all running on Kubernetes. - [Quickstart](https://devx.parthkapoor.me/docs/quickstart): Create your first DevEx REPL, open a terminal, run some code, and see your files persist across restarts — in about five minutes. ## Concepts - [Architecture](https://devx.parthkapoor.me/docs/architecture): The four services behind DevEx — web, core, runner and MCP — what each one owns, and the exact sequence that turns a click into a running container. - [Templates](https://devx.parthkapoor.me/docs/templates): A template is the filesystem and image a new REPL starts from. Here is how they are structured, and the four files you touch to add one. ## Guides - [MCP server](https://devx.parthkapoor.me/docs/mcp): An optional sidecar that exposes a REPL's workspace to AI agents over Model Context Protocol. Today it can read files; write and terminal tools are planned. ## Self-hosting - [Self-hosting](https://devx.parthkapoor.me/docs/self-hosting): Run DevEx workspaces on your own Kubernetes cluster, with the core API on a Docker Swarm host. Any provider; one public node, no cloud load balancer. ## Contributing - [Contributing](https://devx.parthkapoor.me/docs/contributing): How to get DevEx running locally, where each piece of the monorepo lives, and what a good pull request looks like. ## Frequently asked **What is DevEx?** DevEx is an open-source cloud development environment. It schedules a container on Kubernetes and gives you a browser tab with a code editor, a terminal attached to that container's shell, and a file tree backed by persistent storage. **Is DevEx free?** The Free plan costs nothing and needs no card: two workspaces, 125m CPU and 256Mi memory each, 200MB persisted. Self-hosting is free without limits — the source is MIT-licensed and the deployment is documented. **Do my files survive a restart?** Yes. A workspace's filesystem is synced to S3-compatible object storage and restored when the workspace starts again, so stopping one does not lose work. **Can I reach a server running inside a workspace?** Yes. Ports opened inside the container are forwarded to a public URL, so a dev server or an API you started in the terminal is reachable from anywhere without a tunnel. **Can an AI agent use a DevEx workspace?** Partly, today. An optional MCP server can run beside a workspace, so an assistant reads the same files you are editing over the Model Context Protocol. It exposes read_file now; writing files and running commands are planned. **Can I self-host DevEx?** Yes. DevEx needs a Kubernetes cluster, a Redis instance and an S3-compatible bucket. The self-hosting guide covers the manifests, the environment variables and the runner image. ## Project - [Source](https://github.com/parthkapoor-dev/devex): MIT. The whole stack — web app, Go control plane, runner and MCP server. - [Author](https://parthkapoor.me): Parth Kapoor.