Claude.md vs Agents.md

If you're researching Claude.md vs Agents.md, AI DevKit gives your AI coding agents one control plane for setup, console supervision, local-first memory, communication, workflow skills, and verification. AI DevKit works with a variety of AI coding agents and coding environments. This page lists supported environments and explains what AI DevKit installs for each one, so teams can keep setup, skills, memory, and verification consistent across tools.

Before selecting project environments here, complete the once-per-machine setup in Getting Started.

Support levels here describe setup and environment integration. Agent session discovery, agent console, and agent send depend on whether the local agent exposes a detectable running session. See Agent Management and Agent Console for the operational control-plane commands.

Status	Meaning
Ready	Stable setup support for generated files, skills, and documented configuration.
Experimental	Works but may have issues. We're actively testing and improving support.

These environments have stable setup integrations.

What AI DevKit provides:

• AGENTS.md — Agent instructions for Cursor
• .cursor/skills/ — Project-level skills
• ~/.cursor/skills/ — Global skills
• .cursor/rules/ — Editor rules for consistent coding standards

What AI DevKit provides:

• CLAUDE.md — Claude Code workspace instructions and context
• .claude/skills/ — Project-level skills
• ~/.claude/skills/ — Global skills
• .mcp.json — Project-level MCP server configuration

What AI DevKit provides:

• .github/prompts/ — GitHub Copilot custom prompts with VSCode
• .github/skills/ — Project-level skills
• .mcp.json — Project-level MCP server configuration
• ~/.copilot/skills/ — Global skills

What AI DevKit provides:

• GEMINI.md — Context file for providing instructional context to the Gemini model
• .gemini/skills/ — Project-level skills
• ~/.gemini/skills/ — Global skills

What AI DevKit provides:

• AGENTS.md — Codex-specific configuration and context
• .agents/skills/ — Project-level skills
• ~/.codex/skills/ — Global skills
• .codex/config.toml — Project-level MCP server configuration

What AI DevKit provides:

• .grok/skills/ — Project-level skills
• ~/.grok/skills/ — Global skills

What AI DevKit provides:

• AGENTS.md — opencode custom instructions
• .opencode/skills/ — Project-level skills
• ~/.config/opencode/skills/ — Global skills
• opencode.json — Project-level MCP server configuration under the mcp key

What AI DevKit provides:

• .agent/skills/ — Project-level skills
• ~/.gemini/antigravity/skills/ — Global skills

What AI DevKit provides:

• .agents/skills/ — Project-level skills
• ~/.gemini/config/skills/ — Global skills

What AI DevKit provides:

• AGENTS.md — Junie project instructions and context
• .junie/skills/ — Project-level skills
• .junie/mcp/mcp.json — Project-level MCP server configuration
• ~/.junie/skills/ — Global skills

What AI DevKit provides:

• AGENTS.md — Cline project instructions and context
• .cline/skills/ — Project-level skills
• ~/.cline/skills/ — Global skills

What AI DevKit provides:

• AGENTS.md — Devin project instructions and context
• .devin/skills/ — Project-level skills
• .devin/config.json — Project-level MCP server configuration; AI DevKit updates only mcpServers and preserves other Devin config
• ~/.config/devin/skills/ — Global skills

What AI DevKit provides:

• .pi/skills/ — Project-level skills
• ~/.pi/agent/skills/ — Global skills

These environments are under active development. They work, but you may encounter issues.

What AI DevKit provides:

• AGENTS.md — KiloCode configuration for large project handling
• .kilo/skills/ — Project-level skills
• ~/.kilo/skills/ — Global skills
• .kilo/kilo.jsonc — Project-level MCP server configuration under the mcp key

What AI DevKit provides:

• AGENTS.md — AMP configuration for accelerated workflows
• .agents/skills/ — Project-level skills
• ~/.config/agents/skills/ — Global skills

What AI DevKit provides:

• AGENTS.md — Roo Code configuration and context
• .roo/skills/ — Project-level skills
• ~/.roo/skills/ — Global skills
• .roo/mcp.json — Project-level MCP server configuration

When you run ai-devkit init, you can select multiple environments at once:

ai-devkit init

This presents an interactive checklist where you can:

• Spacebar — Select or deselect an environment
• Enter — Confirm your selections
• Select any combination of the supported environments

Your selections are saved in .ai-devkit.json:

{
  "version": "0.69.1",
  "environments": ["cursor", "claude", "github"],
  "phases": ["requirements", "design"],
  "createdAt": "2026-10-10T...",
  "updatedAt": "2026-10-10T..."
}

Want to add another environment after initial setup? Just run:

ai-devkit init

AI DevKit will:

1. Detect your existing environments
2. Ask once before replacing the selected environment templates in interactive mode
3. Ask separately before replacing each existing phase document

In non-interactive mode (--yes), existing environment templates and phase documents are skipped unless you also pass --overwrite. Template-driven initialization overwrites the selected artifacts without prompting.

When an interactively selected environment is already set up, ai-devkit init shows a warning and asks whether to replace the selected environment templates:

Warning: The following environments are already set up: cursor, claude

The following environments are already set up and will be overwritten:
  Cursor, Claude Code

Do you want to continue?

Existing phase documents use a separate per-phase confirmation. This behavior belongs to init; the non-interactive ai-devkit install reconciliation flow does not show a general “existing install artifacts” overwrite prompt.

Want to add support for a new AI environment? We welcome contributions!

1. Create Environment Definition — Add to packages/cli/src/util/env.ts
2. Add Templates — Create templates/env/{code}/ directory
3. Update Documentation — Add to this guide
4. Test Integration — Ensure proper initialization and configuration

See our Contributing Guide for details.

Use AI DevKit to keep Claude.md vs Agents.md consistent across features and teams: one config, one local console, shared memory, and the same verification workflow across supported agents.