Coder

Coder is a self-hosted platform for teams that need cloud development environments and AI coding agents on infrastructure they control. It best suits engineering organizations with infrastructure expertise, security requirements, or varied developer tooling. The trade-off is clear: Coder offers unusual control and access flexibility, but it is less turnkey than a hosted coding service.

Overview

Coder makes development workspaces reproducible through Terraform, with connectivity provided by a secure WireGuard tunnel. A workspace can be opened in VS Code, JetBrains, Cursor, a browser terminal, remote desktop, or SSH, so teams can preserve different editor and access preferences without changing the underlying environment.

Its agent layer is more than an editor add-on. Coder Agents accepts development and research requests in chat or through an API, provisions a workspace for the task, and runs the agent loop in Coder’s control plane. Keeping LLM credentials out of workspaces is a meaningful security boundary for organizations delegating work to models. Coder documents connections to GitHub, GitLab, Bitbucket, and JFrog Artifactory. The company was founded in Austin, Texas, in 2017, and its site displays a SOC 2 Type II Certified badge.

Key features

Terraform-defined environments: Custom workspace configuration and persistent storage support teams that need repeatable, infrastructure-aware setups rather than a fixed template.

Broad workspace access: VS Code, JetBrains, Cursor, web terminal, remote desktop, and SSH cover graphical, browser, and command-line workflows. JetBrains support and GPU workspaces are useful for specialized development, while the choice may add administration compared with a single-purpose browser IDE.

Automatic lifecycle control: Idle workspaces can stop and restart on demand. That can curb the cost and resource use of abandoned environments without forcing developers to recreate them.

Agent delegation: Coder Agents provisions an execution workspace for coding or research tasks submitted by chat or API. The control-plane agent loop separates model credentials from workspace processes.

Model-provider choice: Agents supports Anthropic, OpenAI, Google, Bedrock, and self-hosted endpoints. AI Premium also names model connections for Anthropic, OpenAI, Azure, and AWS Bedrock, giving organizations room to align models with existing procurement or deployment policies.

Team infrastructure: Prebuilds, multiple VCS hosts, GPU workspaces, and integrations with major Git providers make Coder suitable for standardized team environments rather than only personal sandboxes.

Command-line coverage: The Coder CLI runs on Linux, macOS, and Windows, matching the supported desktop operating systems.

Pricing

Community — free. This plan includes unlimited self-hosted workspaces, up to five active Coder Agents, and one platform integration. GitHub and Discord support are included. It is the practical starting point for individuals and small teams able to run their own infrastructure; the five-agent ceiling and single integration constrain larger or more automated deployments.

Premium — custom pricing. Premium adds advanced governance, high availability and workspace proxies, OIDC synchronization, audit logging, and SLA-backed, ticket-based global support. It still lists up to five active Coder Agents, so its value is operational control and support rather than a larger agent allowance. Choose it when reliability, identity integration, and formal support matter more than minimizing spend.

AI Premium — custom pricing. This tier adds self-hosted AI workflows, the Coder Agents API, network-isolated agent workspaces, and model connections for Anthropic, OpenAI, Azure, and AWS Bedrock. It includes unlimited agents, making it the relevant plan for organizations running agent work at scale or requiring isolated execution. The published free plan and trial make evaluation low-risk, but custom-priced tiers require a sales decision.

Platforms

Coder is available as a self-hosted and web platform, with API access and support for Linux, macOS, and Windows through its CLI. Workspaces can be reached from desktop editors, browser tools, remote desktop, or SSH. This is a broad access matrix, but customers remain responsible for operating the underlying infrastructure.

Who it's for

Coder is a strong fit for engineering teams that need customer-controlled infrastructure, Terraform-governed environments, persistent storage, GPU capacity, or strict separation between agents and workspace secrets. It also suits organizations standardizing many developer workflows while retaining editor choice. It is a weaker choice for a solo developer who wants an instantly hosted environment, or for teams unwilling to manage self-hosted services and workspace policy.

Pros and cons

Pros

Infrastructure control: Self-hosting and Terraform configuration let teams align workspaces with their own network and operational requirements.

Editor and protocol flexibility: VS Code, JetBrains, Cursor, browser terminal, remote desktop, and SSH avoid locking developers into one interface.

Credential separation: Running the agent loop in the control plane keeps LLM credentials out of workspaces.

Scalable agent option: AI Premium provides unlimited agents, an API, and network-isolated workspaces.

Cons

Operational burden: Self-hosting provides control but leaves infrastructure operation with the customer.

Community limits: The free plan allows only five active agents and one platform integration.

Premium uncertainty: Premium and AI Premium use custom pricing, making larger deployments harder to budget from the published plans alone.

Alternatives

Choose Replit when a freemium, cross-platform hosted environment with a free daily-usage tier, one live project, and one background task is a better fit than self-hosting.

Playcode is the simpler choice for browser-based projects where the free tier’s three projects or published sites, one collaborator, and required branding are acceptable.

Pick GitHub Codespaces when GitHub-centered development and usage-billed compute are preferable; its listed 2-core and 4-core compute plans cost 0.18 USD and $0.36 a month respectively.

Deepnote suits notebook-focused teams wanting a free tier with up to three editors and five projects.

Arduino Cloud Editor is the better match for Arduino projects with a free allowance of two Things, five variables per Thing, and 25 daily compilations.

Consider RunCode.io for a paid multi-workspace service with a 30-minute inactivity timeout.

Datalore is an alternative for notebook work with a free cloud allocation of 120 CPU S machine hours and 10 GB storage.

Tencent Cloud Cloud Studio is a free or low-cost option with published Chinese-yuan plans and fixed CPU, memory, and disk allocations.

For category browsing, see Coding Playgrounds, Cloud Development Environments, and AI Coding Assistants.

Verdict

Choose Coder if your team wants customer-controlled development infrastructure combined with delegated agent work, broad access methods, and strong environment customization. Look elsewhere if you want a managed, immediately usable coding workspace or predictable published pricing without operating the platform yourself.