Integrations
Connect betterspec to your AI tools and workflows.
Integrations
betterspec uses an adapter architecture to provide native integration with your AI coding tool. No separate plugins or packages are needed — betterspec init scaffolds everything directly into your project.
How It Works
When you run betterspec init, you select your AI tool. betterspec then creates tool-specific files that give your AI assistant:
- Agents — Specialized roles (Planner, Builder, Validator, Archivist) written in each tool's native agent format
- Hooks — Automatic context injection and guardrails using each tool's hook system
- Skills — Reusable workflows installed via agentskills.io or local skill files
- Context files — Tool-specific config files (CLAUDE.md, GEMINI.md, etc.) that orient the AI
This means every supported tool gets first-class support without maintaining separate plugin packages.
Supported Tools
| Tool | Agents | Hooks | Skills | Context |
|---|---|---|---|---|
| OpenCode | .opencode/agents/ | Global plugin | skills/ dir | Plugin injection |
| Claude Code | .claude/agents/ | .claude/hooks/ + settings.json | .claude/agents/ | CLAUDE.md |
| Gemini CLI | .gemini/agents/ | — | — | GEMINI.md |
| Cursor | — | — | — | .cursor/rules/betterspec.mdc |
| Codex | — | — | — | AGENTS.md |
| Generic | — | — | betterspec/skills/ | — |
MCP Server
As an alternative or complement to the adapter approach, betterspec also provides an MCP server that exposes tools and resources over the Model Context Protocol. Any MCP-compatible AI tool can connect to it.
Building Your Own
If you're building an AI tool or workflow system, you can integrate betterspec at two levels:
- Read specs as files — simplest approach, works with any tool
- Use the
@betterspec/coreAPI — programmatic access to config, changes, and scaffolding