Agent Context Explorer (ACE)

Your AI agents are only as good as the context you give them — and that context can be scattered across several sources .cursor/, .claude/, ~/.agents/, and a handful of config files you have to remember exist.

ACE puts it all in one tree. Rules, commands, skills, agent definitions, specs, and registered MCP servers — for every workspace and every agent platform — visible at a glance, always current.

ACE is viewer-only: it scans and displays intentional artifacts without managing them.

What You See

Workspaces view — project-level artifacts for every workspace and added project:

• Cursor — .cursor/rules/*.{mdc,md}, .cursor/commands/*.md, .cursor/skills/*/SKILL.md
• Claude Code — CLAUDE.md, .claude/rules/*.{mdc,md}, .claude/commands/*.md, .claude/skills/*/SKILL.md
• Specs — specs/*/spec.md feature specifications

Global commands and skills (~/.cursor/commands/, ~/.cursor/skills/) are shown alongside workspace artifacts in the Cursor section.

Agents view — user-level agent configuration across all three global roots:

Each root expands into four sections — Agents, Commands, MCP, Skills — in alphabetical order. The MCP section lists every MCP server currently registered for that platform, read straight from ~/.claude.json (Claude Code) or ~/.cursor/mcp.json (Cursor), so you can see what's wired up without opening a config file. It's read-only and stays live: edit either file externally and the list updates within seconds.

The first time ACE detects Claude Code and finds itself unregistered, it offers a one-time prompt to add itself to ~/.claude.json — accept it and ACE (and its context-query tools, see MCP: AI Agent Access below) shows up in that same MCP section going forward.

The view auto-refreshes within seconds when files are added, changed, or removed from any watched path.

Quick Start

1. Open the ACE icon in the Activity Bar.
2. Expand Workspaces to browse Cursor and Claude Code artifacts for your project.
3. Expand Agents to browse your global agent configuration across ~/.cursor/, ~/.claude/, and ~/.agents/.
4. Click any item to open it read-only in your editor.

Multi-Project

Use the + button to add external projects. Each project shows its own rules, commands, skills, and specs independently.

MCP: AI Agent Access

ACE exposes an MCP server so AI agents can query project context on demand.

In Cursor: ACE registers automatically via the VS Code MCP Extension API — no setup needed.

Standalone / fallback: Add to .cursor/mcp.json or ~/.cursor/mcp.json:

{
  "mcpServers": {
    "ace": {
      "command": "node",
      "args": ["<extension-dir>/out/mcp/server.js", "<workspace-root>"]
    }
  }
}

Replace <extension-dir> (e.g. ~/.cursor/extensions/fancy-bread.agent-context-explorer-x.y.z) and <workspace-root> with your actual paths.

Available tools: list_projects, get_project, list_rules, get_rule, list_commands, get_command, list_skills, get_skill, list_agents, get_agent, list_specs, get_spec

Requirements

VS Code or Cursor ^1.105.0 (see engines.vscode in package.json).

Troubleshooting

License

MIT — see LICENSE.