# Platforms

Assemble generates native configuration files for 21 AI coding platforms. Each platform reads its own config format — Assemble handles the translation.

## The adapter pattern

Assemble uses an **adapter pattern**: the same agent definitions and workflows are translated into platform-specific formats. You define your team once, and Assemble generates the right files for each platform.

```
Agent definitions (.md)  →  Adapter  →  .cursorrules

→  Adapter  →  CLAUDE.md

→  Adapter  →  .github/copilot-instructions.md

→  Adapter  →  .windsurfrules

→  ...
```

## Supported IDE platforms (16)

| Platform | Config file | Notes |
| --- | --- | --- |
| Cursor | `.cursorrules` | Most popular AI-first IDE |
| Windsurf | `.windsurfrules` | Codeium-powered IDE |
| VS Code + Copilot | `.github/copilot-instructions.md` | Most widely used IDE |
| JetBrains AI | `.junie/guidelines.md` | IntelliJ, WebStorm, PyCharm, etc. |
| Zed | `.zed/prompt.md` | High-performance editor |
| Trae | `.trae/rules` | Bytedance AI editor |
| Void | `.void/rules.md` | Open-source AI editor |
| PearAI | `.pearai/rules.md` | Fork of VS Code with AI |
| Melty | `.melty/rules.md` | Collaborative AI editor |
| Codex | `codex.md` | OpenAI’s coding agent |
| Bolt | `.bolt/prompt` | StackBlitz AI editor |
| Lovable | `.lovable/rules.md` | AI web app builder |
| Replit | `.replit/agent/system_prompt.md` | Cloud IDE |
| Cline | `.clinerules` | VS Code AI extension |
| Roo Code | `.roo/rules.md` | VS Code AI extension |
| Tabnine | `.tabnine/rules.md` | AI code completion |

## Supported CLI platforms (5)

| Platform | Config file | Notes |
| --- | --- | --- |
| Claude Code | `CLAUDE.md` | Anthropic’s CLI coding agent |
| Aider | `.aider/conventions.md` | Terminal-based pair programming |
| Continue | `.continue/rules.md` | Open-source AI assistant |
| AugmentCode | `.augment/rules.md` | Enterprise AI coding |
| Amp | `amp.md` | Sourcegraph’s coding agent |

## How generation works

When you run `npx cohesiumai-assemble` or `npx cohesiumai-assemble generate`:

1. **Read**`.assemble.yaml` for platform list and configuration
2. **Load** agent definitions from `.assemble/agents/`
3. **Load** workflow definitions from `.assemble/workflows/`
4. **For each platform**: apply the adapter to produce the native config file
5. **Write** the generated files to the project root

### What gets included in each file?

Each generated file contains:

- **Orchestrator prompt** (Jarvis routing logic)
- **Agent definitions** (filtered by relevance to the platform)
- **Workflow definitions** (adapted to the platform’s capabilities)
- **Team configuration** (roles, domains, collaboration rules)
- **Project-specific context** (from `.assemble.yaml`)

### Platform-specific adaptations

Different platforms have different capabilities. Assemble adapts:

| Feature | Cursor | Claude Code | Copilot |
| --- | --- | --- | --- |
| File-based context | `.cursorrules` | `CLAUDE.md` \+ `.claude/rules/` | `.github/copilot-instructions.md` |
| Sub-agent delegation | Limited | Agent tool | Limited |
| Multi-file support | Single file | Directory tree | Single file |
| Max context size | ~6000 tokens | Large | ~4000 tokens |

## Updating platform files

After changing your agent definitions or workflows:

```
npx cohesiumai-assemble generate
```

This regenerates all platform files while preserving your `.assemble.yaml` configuration.

## Adding a new platform

As new AI coding tools emerge, Assemble adds adapters. To request a new platform:

1. Open an issue on [GitHub](https://github.com/CohesiumAI/assemble/issues)
2. Include the platform name and its config file format
3. The community will build and test the adapter