Coder: Secure Cloud Environments for Developers and AI Agents
Coder provisions secure remote dev environments for engineers and AI agents. Go, AGPL-3.0, 15,307 stars, with self-hosted and cloud editions at coder.com.
TL;DR:
coder/coderis the Go, AGPL-3.0 platform behind coder.com that provisions secure remote development environments for human engineers and AI agents alike. 15,307 stars on GitHub, both self-hosted and managed editions.
What it is
coder/coder describes itself as “Secure environments for developers and their agents” — a short tagline that covers both the original use case (remote dev workspaces in the browser) and the newer one (workspaces purpose-built for AI coding agents). The repository is written in Go, has 15,307 stars, and is published under the GNU Affero General Public License v3.0. The homepage field is https://coder.com, where the project ships both an open-source self-hosted edition and a managed cloud product. The GitHub topic list is short and on-message: agents, dev-tools, development-environment, go, golang. The repository was created on 2021-12-22 and uses main as its default branch.
The shape is a control plane plus workspace templates. You define what a workspace looks like — container image, IDE, dotfiles, secret mounts — and Coder provisions it on your infrastructure or in the managed cloud. Both humans and agents connect through the same surface, which is why the “agents” topic sits next to “development-environment” in the list.
What it does well
The interesting tension Coder solves is that remote development environments used to be a people-only story, and the same primitives now need to serve non-human clients. The README and topic list together signal how that is wired.
- Workspaces for humans and agents from the same control plane. The
agentstopic next todevelopment-environmentis the giveaway: the same template language describes both interactive IDEs and headless agent runs. - Bring-your-own infrastructure. Coder is AGPL-3.0 and self-hostable, which means a workspace can run on your own Kubernetes cluster, on a single VM, or on coder.com’s managed plane. Your code and credentials stay where you put them.
- IDE-agnostic surface. Workspaces serve VS Code, JetBrains, and the terminal through a single protocol layer, and the same applies to agents that only need shell access.
- Template-driven provisioning. Workspaces are described in
coder/coderas declarative templates, which makes spinning up a reproducible dev environment acoder createinstead of a wiki page. - Enterprise security primitives. The “secure environments” framing covers SSH and identity integration, and the AGPL-3.0 license means any hosted fork has to publish changes.
The 15,307-star count at 478 stars today shows a project that has settled into a stable niche rather than chasing the hype cycle.
How to try it
The fastest path is the self-hosted install script, which provisions a single-node Coder deployment you can poke at.
# Install the Coder CLI
curl -L https://coder.com/install.sh | sh
# Stand up a single-node deployment locally
coder server
# Create a workspace from a built-in template
coder create --template docker
For a managed setup, the homepage at https://coder.com walks through creating a workspace backed by coder.com’s cloud.
# Log in to the managed plane
coder login https://example.coder.com
# List available templates
coder templates list
When it fits, and when it doesn’t
Coder is the right answer when you have engineers (or agents) who should not be running on a laptop, when reproducibility matters more than local IDE conveniences, or when you want a single control plane for both kinds of clients. It is a strong fit for security-sensitive shops that need to keep source on a private cluster, and for teams shipping AI agents that need a clean, reproducible shell. It is less attractive for a single developer who is happy with a local docker-compose dev setup and does not need cross-machine workspaces, or for very small teams where the operational overhead of running Coder itself outweighs the consistency win. The AGPL-3.0 license is also worth a sanity check before embedding Coder into a commercial product.
The 15,307-star count, the five-year-old repository, and the dual human/agent positioning are the signals worth weighing: a mature, self-hostable platform whose audience already includes teams that ship coding agents.
Source and Accuracy Notes
- Project page: https://github.com/coder/coder
- Source repository:
coder/coder(verified via GitHub APIhtml_url) - License: AGPL-3.0
- Stars at time of writing: 15,307
- Language: Go
- Topics:
agents,dev-tools,development-environment,go,golang - Homepage: https://coder.com
- Created: 2021-12-22
Related Posts
ai-setup
Recall – Persistent Memory for Claude Code via MCP Hooks
Recall gives Claude Code a permanent memory store that survives session restarts and context compaction. Four hooks capture and restore context automatically — with cloud SaaS or self-hosted options.
2/28/2026
dev-tools
Automotive Skills Suite for AI Engineering
Evaluate Automotive Skills Suite for APQP, ASPICE, HARA, safety-plan, and DIA workflows with setup notes, governance risks, and SME review guidance.
5/28/2026
dev-tools
awesome-agentic-ai-zh Roadmap Guide
Explore awesome-agentic-ai-zh as a Chinese agentic AI learning roadmap, with setup notes, track selection, study workflow, and evaluation guidance.
5/28/2026