Opencode

OpenCode plugin for Claude Code

Use OpenCode from inside Claude Code for code reviews or to delegate tasks to OpenCode.

This plugin is for Claude Code users who want an easy way to start using OpenCode (any provider, any model — Anthropic, OpenAI, OpenRouter, local) from the workflow they already have. It mirrors the openai/codex-plugin-cc plugin's surface, with the delegate swapped from Codex to OpenCode.

What You Get

• /opencode:review for a focused review of your current work
• /opencode:adversarial-review for a steerable challenge review
• /opencode:rescue, /opencode:status, /opencode:result, and /opencode:cancel to delegate work and manage background jobs

Requirements

• A working opencode CLI with at least one provider configured (opencode auth login).
  • Usage will draw on whichever provider OpenCode is configured to use.
• Node.js 18.18 or later

Install

Add the marketplace in Claude Code:

/plugin marketplace add mubeda/opencode-plugin-cc

Install the plugin:

/plugin install opencode@opencode-cc

Reload plugins:

/reload-plugins

Then run:

/opencode:setup

/opencode:setup will tell you whether OpenCode is ready. If OpenCode is missing and npm is available, it can offer to install OpenCode for you.

If you prefer to install OpenCode yourself, use:

npm install -g opencode-ai

If OpenCode is installed but has no provider configured yet, run:

!opencode auth login

After install, you should see:

• the slash commands listed below
• the opencode:opencode-rescue subagent in /agents

One simple first run is:

/opencode:review --background
/opencode:status
/opencode:result

Updating an installed version

Refresh the marketplace catalog, then update the plugin:

/plugin marketplace update opencode-cc
/plugin update opencode@opencode-cc

Reload plugins to pick up the new version:

/reload-plugins

Usage

/opencode:review

Runs an OpenCode review on your current work using a provider-neutral review prompt. Output is whatever the configured model produces in Markdown.

[!NOTE] Reviews of multi-file changes might take a while. It's generally recommended to run them in the background.

Use it when you want:

• a review of your current uncommitted changes
• a review of your branch compared to a base branch like main

Use --base <ref> for branch review. It also supports --wait and --background. It is not steerable and does not take custom focus text. Use /opencode:adversarial-review when you want to challenge a specific decision or risk area.

Examples:

/opencode:review
/opencode:review --base main
/opencode:review --background

This command is read-only and will not perform any changes itself. The companion runs OpenCode with --dangerously-skip-permissions, but the review prompt explicitly tells the model not to modify files.

/opencode:adversarial-review

Runs a steerable review that questions the chosen implementation and design.

It can be used to pressure-test assumptions, tradeoffs, failure modes, and whether a different approach would have been safer or simpler.

It uses the same review target selection as /opencode:review, including --base <ref> for branch review. It also supports --wait and --background. Unlike /opencode:review, it can take extra focus text after the flags.

Use it when you want:

• a review before shipping that challenges the direction, not just the code details
• review focused on design choices, tradeoffs, hidden assumptions, and alternative approaches
• pressure-testing around specific risk areas like auth, data loss, rollback, race conditions, or reliability

Examples:

/opencode:adversarial-review
/opencode:adversarial-review --base main challenge whether this was the right caching and retry design
/opencode:adversarial-review --background look for race conditions and question the chosen approach

This command is read-only. It does not fix code.

/opencode:rescue

Hands a task to OpenCode through the opencode:opencode-rescue subagent.

Use it when you want OpenCode to:

• investigate a bug
• try a fix
• continue a previous OpenCode task
• take a faster or cheaper pass with a smaller model

[!NOTE] Depending on the task and the model you choose these tasks might take a long time and it's generally recommended to force the task to be in the background or move the agent to the background.

It supports --background, --wait, --resume, and --fresh. If you omit --resume and --fresh, the plugin can offer to continue the latest rescue thread for this repo.

Examples:

/opencode:rescue investigate why the tests started failing
/opencode:rescue fix the failing test with the smallest safe patch
/opencode:rescue --resume apply the top fix from the last run
/opencode:rescue --model openrouter/anthropic/claude-sonnet-4.6 --effort high investigate the flaky integration test
/opencode:rescue --background investigate the regression

You can also just ask for a task to be delegated to OpenCode:

Ask OpenCode to redesign the database connection to be more resilient.

Notes:

• if you do not pass --model or --effort, OpenCode resolves them from your opencode.json (or the global default).
• --model takes the OpenCode-style provider/model identifier. Use opencode models to list what's available locally.
• follow-up rescue requests can continue the latest OpenCode session in the repo.

/opencode:status

Shows running and recent OpenCode jobs for the current repository.

Examples:

/opencode:status
/opencode:status task-abc123

Use it to:

• check progress on background work
• see the latest completed job
• confirm whether a task is still running

/opencode:result

Shows the final stored OpenCode output for a finished job. When available, it also includes the OpenCode session ID so you can reopen that run directly in OpenCode with opencode run -s <session-id>.

Examples:

/opencode:result
/opencode:result task-abc123

/opencode:cancel

Cancels an active background OpenCode job.

Examples:

/opencode:cancel
/opencode:cancel task-abc123

/opencode:setup

Checks whether OpenCode is installed and authenticated. If OpenCode is missing and npm is available, it can offer to install OpenCode for you.

Typical Flows

Review Before Shipping

/opencode:review

Hand A Problem To OpenCode

/opencode:rescue investigate why the build is failing in CI

Start Something Long-Running

/opencode:adversarial-review --background
/opencode:rescue --background investigate the flaky test

Then check in with:

/opencode:status
/opencode:result

OpenCode Integration

The OpenCode plugin spawns the local opencode CLI per task with --format json and parses the resulting event stream:

opencode run --format json --dir <repo> --dangerously-skip-permissions \
  [-m <provider/model>] [--variant <effort>] [-s <session-id>] "<prompt>"

The companion captures the sessionID from the first event, tracks each job in .claude/plugins/data/<plugin>/state/<workspace>/jobs/, and forwards the assistant's final message verbatim to Claude Code.

Common Configurations

If you want to change the default model or provider that gets used by the plugin, configure that inside your user-level or project-level opencode.json. For example to always use Claude Sonnet on OpenRouter for a specific project you can add the following to an opencode.json at the root of the directory you started Claude in:

{
  "model": "openrouter/anthropic/claude-sonnet-4.6"
}

OpenCode's configuration resolves:

• user-level config in ~/.config/opencode/opencode.json
• project-level overrides in opencode.json at the repo root

Check out the OpenCode docs for more configuration options.

Moving The Work Over To OpenCode

Delegated tasks can be directly resumed inside OpenCode by running opencode run -s <session-id> with the specific session ID you received from running /opencode:result or /opencode:status, or by using opencode session list and picking one.

This way you can review the OpenCode work or continue the work there.

FAQ

Do I need a separate account for this plugin?

No. The plugin uses whichever provider you have already authenticated with opencode auth login. If OpenCode runs for you in the terminal, the plugin should run for you here.

Does the plugin use a separate OpenCode runtime?

No. This plugin delegates through your local OpenCode CLI on the same machine.

That means:

• it uses the same OpenCode install you would use directly
• it uses the same local authentication and provider configuration
• it uses the same repository checkout and machine-local environment

Will it use the same OpenCode config I already have?

Yes. If you already use OpenCode, the plugin picks up the same configuration.

Why does the companion pass --dangerously-skip-permissions?

OpenCode otherwise prompts for permission on each tool call, which would hang an unattended companion run. The review commands counter this by telling the model not to modify files; rescue commands trust the user's instructions.

How does this relate to the Codex plugin?

This plugin is a port of openai/codex-plugin-cc (Apache-2.0). The command surface, job state machine, and orchestration logic are derived from that codebase. The bridge layer that talks to the delegate CLI is rewritten for OpenCode.