Skip to main content

Ecosystem

Providers

Providers connect OpenCode to LLM backends. The ecosystem includes:

  • Cloud providers: Anthropic, OpenAI, Google, AWS Bedrock, Azure, Groq, Together AI, Fireworks, Mistral, Cohere
  • Local providers: Ollama, LM Studio, llama.cpp, vLLM, text-generation-webui
  • Custom providers: Any OpenAI-compatible API via @ai-sdk/openai-compatible

Discover and configure providers through the Models.dev directory.

Plugins

Community and official plugins extend OpenCode with new capabilities:

PluginDescription
opencode-helicone-sessionObservability and analytics via Helicone
@my-org/gitlab-toolsGitLab MR, issue, and pipeline management
opencode-jiraJira ticket lookup and creation
opencode-slackSlack message search and send
opencode-dockerDocker container and image management

See the Plugins guide for authoring and installing plugins.

Skills

Skills are stored in .opencode/skills/ or ~/.config/opencode/skills/ and are loaded automatically. Each skill is a directory with a SKILL.md file and optional resource files.

# Create a skill
mkdir -p .opencode/skills/security-review
# .opencode/skills/security-review/SKILL.md

Review code for common security vulnerabilities including:
- SQL injection
- XSS
- CSRF
- Insecure deserialization
- Hardcoded secrets

## Usage

Run this skill when asked to perform a security audit.

Agents

Agents are specialized assistants configured for specific modes and domains. The ecosystem includes pre-built agent definitions:

  • Build agent: Default coding assistant with full tool access
  • Plan agent: Architecture and design focused, uses plan mode
  • Custom agents: Define your own in opencode.json
{
"agent": {
"security-expert": {
"mode": "plan",
"model": "anthropic/claude-sonnet-4",
"system": "You are a security engineering expert. Analyze code for vulnerabilities.",
"temperature": 0.1,
"permission": {
"read_file": "allow",
"run_command": "deny"
}
}
}

tip

You can define per-agent permissions, models, and system prompts for specialized workflows.

MCP Servers

The Model Context Protocol (MCP) servers connect OpenCode to external tools and data sources:

  • Filesystem: Access local files with restricted paths
  • GitHub: Repository management, PRs, issues
  • PostgreSQL: Database schema inspection and querying
  • Puppeteer: Browser automation and web scraping
  • Sentry: Error tracking and performance monitoring

Configure MCP servers in opencode.json:

{
"mcp": {
"filesystem": {
"type": "stdio",
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-filesystem", "/workspace"]
},
"github": {
"type": "url",
"url": "https://mcp.github.com/sse",
"enabled": true
}
}
}

Community

Contribution Opportunities

There are many ways to contribute to the OpenCode ecosystem:

Publish a Plugin

Package your custom tools and hooks as an npm package. Plugins are the primary way to share integrations with the community. Follow the SDK guide for plugin development.

Add a Provider

If you use a model provider that isn't supported, create a custom provider configuration and share it. Submit a pull request to add it to the built-in provider list.

Create a Skill

Share repeatable workflows as skills. Skills are plain markdown files with structured instructions — easy to create and distribute.

Improve Documentation

Documentation contributions are always welcome. Submit PRs against the docs repository to fix errors, add examples, or write new guides.

Report Issues

Found a bug or missing feature? Open an issue on GitHub with a clear description and reproduction steps.

Join the Community

Participate in Discord discussions, answer questions, and help other users. The community grows stronger with every contribution.

tip

New to the ecosystem? Start with a custom skill or agent definition before publishing a plugin.

Architecture Overview

+-----------------+
| OpenCode CLI |
+--------+--------+
|
+----------------+----------------+
| | |
+------v------+ +-----v-----+ +------v------+
| TUI App | | Server | | Run (CLI) |
+------+------+ +-----+-----+ +------+------+
| | |
+--------+------+----------------+
|
+--------v--------+
| Agent Engine |
+--------+--------+
|
+------------+------------+
| | |
+----v----+ +---v----+ +---v----+
| Models | | Tools | | MCP |
+----+----+ +---+----+ +---+----+
| | |
+----v----+ | +----v----+
|Providers| | |External |
+---------+ | |Services |
| +---------+
+------v------+
| Plugins |
+-------------+