OpenCode

What it does

Writes code

YesAuthoring and running code from a prompt is the product's stated job across all four surfaces.OpenCode is an open source agent that helps you write and run code with any AI model.opencode.ai ↗

Debugs and fixes

YesAsked to fix an issue, it works in a new branch and submits a PR with the changes.Fix and implement: Ask OpenCode to fix an issue or implement a feature. And it will work in a new branch and submits a PR with all the changes.opencode.ai/docs/github ↗

Reviews code

YesInvoked on a review comment it receives the file path, line numbers and diff context, and PR-open events can trigger it automatically.Mention /opencode or /oc while reviewing code. OpenCode receives file path, line numbers, and diff context.opencode.ai/docs/github ↗

Whole-codebase work

PartlyA dedicated read-only subagent explores whole codebases and LSPs load automatically, but no repository-wide migration claim is published.A fast, read-only agent for exploring codebases. Cannot modify files. Use this when you need to quickly find files by patterns, search code for…opencode.ai/docs/agents ↗

Ships it

PartlyIt creates branches and opens pull requests inside your CI runner; no deploy or release step is documented.Mention /opencode or /oc in your comment. OpenCode reads context and can create branches, open PRs, or reply.opencode.ai/docs/github ↗

Not established: Writes tests

How much it does unattended

Runs autonomously

YesAn auto mode approves every request that is not explicitly denied, and the same flag works on the non-interactive run command.Start OpenCode with --auto to automatically approve permission requests that are not explicitly denied.opencode.ai/docs/permissions ↗

Multi-agent

YesPrimary agents delegate to subagents, and the vendor also advertises starting multiple agents in parallel on one project.Subagents are specialized assistants that primary agents can invoke for specific tasks.opencode.ai/docs/agents ↗

Agent permissions

YesEach action resolves to allow, ask or deny, with per-command and per-path granularity plus guards for external directories and repeated calls.OpenCode uses the permission config to decide whether a given action should run automatically, prompt you, or be blocked.opencode.ai/docs/permissions ↗

Which models it runs on

Claude

YesClaude is named on the hero as a connectable provider, and Claude models are sold through the vendor's own Zen gateway.Free models included or connect any model from any provider, including Claude, GPT, Gemini and more.opencode.ai ↗

GPT

YesGPT works via an OpenAI key, a Zen endpoint, or by signing in with an existing ChatGPT Plus/Pro subscription.Yes, OpenCode supports subscription plans from all major providers. You can use your Claude Pro/Max, ChatGPT Plus/Pro, or GitHub Copilot…opencode.ai ↗

Open models

YesOpen-weight models run locally through Ollama, LM Studio or llama.cpp, and the Go subscription is built specifically around open-source models.You can configure opencode to use local models through Ollama.opencode.ai/docs/providers ↗

Model choice

YesYou choose the provider and model; the vendor's own gateway is explicitly optional.OpenCode uses the AI SDK and Models.dev to support 75+ LLM providers and it supports running local models.opencode.ai/docs/providers ↗

Where you use it

In your editor

YesAn extension covers VS Code and its forks, and ACP support puts it inside Zed and JetBrains IDEs.OpenCode integrates with VS Code, Cursor, or any IDE that supports a terminal.opencode.ai/docs/ide ↗

On the command line

YesThe binary opens a TUI by default and also accepts commands for scripted, non-interactive runs.The OpenCode CLI by default starts the TUI when run without any arguments.opencode.ai/docs/cli ↗

In your pipeline

YesIt executes inside your GitHub Actions runner, and a parallel integration runs it in GitLab CI or GitLab Duo.OpenCode integrates with your GitHub workflow. Mention /opencode or /oc in your comment, and OpenCode will execute tasks within your GitHub Actions…opencode.ai/docs/github ↗

In a browser

Yesopencode web starts a server bound to 127.0.0.1 and opens the browser UI, so the web surface is one you run rather than a vendor-hosted service.OpenCode can run as a web application in your browser, providing the same powerful AI coding experience without needing a terminal.opencode.ai/docs/web ↗

Whose machine it runs on

Self-hosted

YesThe agent runs entirely on your own machine or runners, talking directly to whichever provider or internal gateway you point it at.All processing happens locally or through direct API calls to your AI provider.opencode.ai/docs/enterprise ↗

Open source

YesMIT-licensed with a public repository that takes outside issues and pull requests.Yes, OpenCode is fully open source. The source code is public on GitHub under the MIT License, meaning anyone can use, modify, or contribute to its…opencode.ai ↗

What it costs to run

How it meters

YesZen meters tokens per 1M with input, output and cached-read rates published per model; Go is a flat monthly subscription and Enterprise is per seat.We support a pay-as-you-go model. Below are the prices per 1M tokens.opencode.ai/docs/zen ↗

Free tier

YesThe agent itself costs nothing and bundles free models; you only pay if you attach a paid provider.OpenCode is 100% free to use. It also comes with a set of free models. There might be additional costs if you connect any other provider.opencode.ai ↗

API access

YesA headless HTTP server with an OpenAPI spec, plus a published SDK, both in the free open-source binary.The opencode serve command runs a headless HTTP server that exposes an OpenAPI endpoint that an opencode client can use.opencode.ai/docs/server ↗

MCP server

NoMCP runs inbound only: external MCP servers become tools inside OpenCode, and no MCP endpoint exposing OpenCode itself is documented.You can add external tools to OpenCode using the Model Context Protocol, or MCP. OpenCode supports both local and remote servers.opencode.ai/docs/mcp-servers ↗

Bring your own key

YesAny provider works from your own API keys, and Zen additionally accepts your own OpenAI or Anthropic keys billed by the provider.With OpenCode you can use any LLM provider by configuring their API keys.opencode.ai/docs ↗

Buying it for a team

A company can buy it

YesEnterprise is sold per seat by quote, and Zen workspaces carry team members, roles and shared billing.We use a per-seat model for OpenCode Enterprise. If you have your own LLM gateway, we do not charge for tokens used.opencode.ai/docs/enterprise ↗

Seat model

PartlyPer-seat pricing is confirmed but the seat rate is quote-only, and no seat floor or ceiling is published.We offer per-seat enterprise pricing. If you have your own LLM gateway, we do not charge for tokens used.opencode.ai/enterprise ↗

Pooled budget

YesA Zen workspace draws on one balance with a workspace-wide monthly cap plus optional per-member caps.You can also set a monthly usage limit for the entire workspace and for each member of your team.opencode.ai/docs/zen ↗

Admin controls

YesPermissions are set in config globally or per tool, so an operator can deny edits or shell commands outright.You can set permissions globally (with *), and override specific tools.opencode.ai/docs/permissions ↗

Single sign-on

PartlySSO arrives only through the enterprise central config, and the vendor does not say whether that means SAML, OIDC or both.Through the central config, OpenCode can integrate with your organization’s SSO provider for authentication.opencode.ai/docs/enterprise ↗

Not established: Audit log

What happens to your code

Opt out of training

PartlyZen providers are zero-retention and non-training by default, but the vendor lists named exceptions where free and contributor models do train on your prompts.Our providers follow a zero-retention policy and do not use your data for model training, with the following exceptions:opencode.ai/docs/zen ↗

Data residency

PartlyGateway inference is US-only with no other region offered; your code stays on your machine because the agent stores nothing.All Zen models are hosted in the US.opencode.ai/zen ↗

Getting out

PartlyBilling can be switched off at any time and the remaining balance stays spendable.Yes, you can disable billing at any time and use your remaining balance.opencode.ai/zen ↗

Certifications

PartlyA SOC 2 Type 2 report is available under NDA through the Trust Center, whose framework list shows SOC 2 Type 1 and Type 2 only.Our Trust Center has everything: SOC 2 Type 2 report, security policies, subprocessor list, and answers to common security questions.opencode.ai/enterprise ↗