Coder

Coder is a self-hosted platform that combines standardized development workspaces, a model-agnostic AI coding agent, and centralized governance with audit logging. It targets enterprise platform and security teams that need AI-assisted development to run on their own infrastructure under their own policies.

Coder is a self-hosted development infrastructure platform built around three layers: standardized cloud workspaces, an AI agent layer, and governance tooling that sits on top of both. The core argument of the product is that developers already operate inside policy, review and audit frameworks, while the AI assistants they use often do not. Coder closes that gap by moving agent activity onto infrastructure the organization owns, where permissions, logging and model access can be dictated centrally. In practice, teams provision development environments from a shared configuration rather than letting each engineer assemble a local machine. Those self-hosted workspaces give developers and agents the same consistent starting point, which removes the configuration drift that makes debugging and onboarding slow. On top of that, Coder Agents provides a native AI coding agent that can be pointed at whichever model the company chooses, and Coder AI Governance acts as a gateway: it centralizes model access, applies policy to environments, and records usage so that every AI request can be observed and audited after the fact. The intended buyer is an enterprise platform, DevOps or security engineering group rather than an individual developer. The reference list (Dropbox, Palantir, J.B. Hunt, Discord, Square, Mercedes, KKR, PsiQuantum and a U.S. Department of Defense software group) points to organizations with compliance obligations and large engineering headcounts, and the published case studies emphasize measurable outcomes such as faster onboarding and steep reductions in VDI cost. Compared with hosted AI coding assistants, Coder competes on control rather than on model quality. The tradeoff is that it is infrastructure: someone has to run it, and the value only materializes once workspaces and policies are properly designed. Teams that just want a chat assistant in their editor are not the audience. Teams that want agentic coding to pass a security review are.