Skip to main content
The OpenCode Agent Catalog provides 69 specialized AI agents covering languages, frameworks, infrastructure, security, AI/ML, and business domains. Each agent follows a standardized expert format with identity, decision trees, examples, and quality gates.

​
Agent Format

Every agent is a markdown file with YAML frontmatter containing:
  • Description — What the agent specializes in
  • Mode — Where it can run (primary, subagent, or all)
  • Permission — Required tool access (e.g., readonly, edit, full)
  • Tags — Searchable keywords
The agent body follows a 4-section structure:
  1. Identity — Role definition, expertise scope, and version context
  2. Decisions — IF/THEN decision trees for common scenarios
  3. Examples — Code samples showing best practices
  4. Quality Gate — Validation criteria before delivering work

​
Quality Scoring

All agents are evaluated on 8 dimensions (1-5 scale):
DimensionWhat’s MeasuredExcellent (5/5)
FrontmatterRequired metadata fieldsAll 3 fields present
IdentityClear role paragraph50-300 words
DecisionsDecision trees with IF/THEN≥ 5 rules
ExamplesCode samples≥ 3 examples
Quality GateValidation criteria≥ 5 criteria
ConcisenessLine count and filler ratio70-120 lines, ≤ 3% filler
No banned sectionsAbsence of deprecated sections0 legacy sections
Version pinningReferences to specific versionsVersion + year present
Pass threshold: Average ≥ 3.5 AND no dimension < 2 Current stats: 69 agents · 4.59/5 average · 100% pass rate · 49 Excellent, 20 Good

​
Categories

Agents are organized into 10 categories:

Languages

11 agents for TypeScript, Python, Go, Rust, Java, and more

AI & Machine Learning

9 agents for AI systems, ML pipelines, and LLM architecture

Web & Mobile

9 agents for React, Next.js, Vue, Angular, and mobile development

Data & API

6 agents for API design, databases, GraphQL, and Redis

DevOps & Infrastructure

9 agents for Docker, Kubernetes, Terraform, AWS, and CI/CD

DevTools

8 agents for code review, debugging, testing, and refactoring

Security

5 agents for security audits, penetration testing, and compliance

MCP

2 agents for Model Context Protocol server development

Business & Management

6 agents for product management, Scrum, and UX research

Documentation

4 agents for technical writing, API docs, and diagrams

​
Installation

Install agents using the OpenCode CLI: 
# Install a specific agent
npx github:dmicheneau/opencode-template-agent install typescript-pro

# Install a category
npx github:dmicheneau/opencode-template-agent install --category languages

# Install a pack (curated bundles)
npx github:dmicheneau/opencode-template-agent install --pack backend
Or use the interactive TUI: 
npx github:dmicheneau/opencode-template-agent tui

​
Agent Modes

  • primary — Designed to be your main assistant for a session
  • subagent — Specialized expert invoked via @category/name
  • all — Works in both modes

​
FAQ

Use the format @category/agent-name in OpenCode:
@languages/typescript-pro Review this API client code
@security/security-auditor Audit this authentication flow
Yes. OpenCode supports multi-agent workflows. For example:
@web/react-specialist Build the component
@devtools/test-automator Write tests for it
@devtools/code-reviewer Review both
  • Agents are individual specialists (e.g., typescript-pro)
  • Packs are curated bundles of agents for common workflows (e.g., backend includes postgres-pro, redis-specialist, api-architect, etc.)
Agents are curated manually and updated when new best practices, framework versions, or tooling evolves. The quality scoring system ensures all agents maintain high standards.
Yes. Create a markdown file in .opencode/agents/<category>/<name>.md following the 4-section format. See the source repository for templates.

Build docs developers (and LLMs) love

Get started for freeTalk to us