Home - JoelBondoux/AtlasMind GitHub Wiki
AtlasMind Wiki
Welcome to the AtlasMind wiki — a comprehensive guide to the multi-agent AI orchestrator for VS Code.
What is AtlasMind?
AtlasMind turns VS Code into a full agentic development environment. Instead of a single chatbot, you get an orchestrator that picks the right agent, the right model, and the right tools for every task — then tracks cost and remembers decisions across sessions.
Key highlights:
- Multi-agent — define specialised agents (architect, tester, security reviewer, etc.) and let the orchestrator route automatically
- Multi-provider model routing — Claude, GPT, Gemini, DeepSeek, Mistral, z.ai, GitHub Copilot, or local models. Budget and speed preferences steer selection
- 26 built-in skills — file I/O, git operations, diagnostics, code navigation, test running, web fetch, and more
- Long-term project memory (SSOT) — decisions, architecture notes, domain knowledge, and lessons persist in a structured memory folder
- Autonomous project planner — decompose goals into parallel subtasks, preview impact, gate with approvals, and review results
- Real-time cost tracking — per-session spend with budget guardrails
Quick Navigation
| Page | Description |
|---|---|
| Getting Started | Install, configure, and run your first command |
| Architecture | System design, core services, data flow |
| Chat Commands | All @atlas slash commands with usage examples |
| Agents | Built-in and custom agent definitions |
| Skills | 26 built-in skills, custom skill import, MCP tools |
| Model Routing | Provider setup, budget/speed modes, routing algorithm |
| Memory System | SSOT folder structure, retrieval, bootstrapping, import |
| Project Planner | Autonomous task decomposition and parallel execution |
| Tool Execution | Approval gating, terminal allow-list, checkpoints, webhooks |
| Configuration | All extension settings with defaults |
| Funding and Sponsorship | How to support ongoing AtlasMind development |
| Security | Security model, boundaries, vulnerability reporting |
| Contributing | Dev setup, coding conventions, PR checklist |
| FAQ | Common questions and troubleshooting |
| Comparison | How AtlasMind compares to Claude Code, Cursor, Copilot, Aider, and Open Hands |
| Changelog | Version history and release notes |
How It Compares
| Capability | AtlasMind | Claude Code | Cursor | Copilot | Aider | Open Hands |
|---|---|---|---|---|---|---|
| Runs inside VS Code | ✅ | ✅ | ✅ (fork) | ✅ | ❌ (terminal) | ❌ (browser/GUI) |
| Multiple AI agents | ✅ | ✅ | ❌ | ⚠️ sessions and agent types | ❌ | ✅ |
| Custom agent definitions | ✅ | ✅ | ❌ | ✅ | ❌ | ⚠️ limited |
| Multi-provider model routing | ✅ | ⚠️ third-party providers | ✅ | ⚠️ built-in and third-party agents | ✅ | ✅ |
| Budget-aware model selection | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ |
| Long-term project memory | ✅ (SSOT) | ⚠️ (CLAUDE.md + memory) | ❌ | ⚠️ custom instructions/context | ❌ | ❌ |
| Memory security scanning | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ |
| Built-in skills / tools | 26 | ~15 | ~10 | ~8 | ~6 | ~20 |
| MCP server integration | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ |
| Custom skill import | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ |
| Autonomous project planner | ✅ | ⚠️ agent workflows | ⚠️ plan mode | ⚠️ plan agent | ❌ | ✅ |
| Per-tool approval gating | ✅ | ✅ | ✅ | ⚠️ varies by agent/tool | ✅ | ❌ |
| Real-time cost tracking | ✅ | ❌ | ❌ | ❌ | ⚠️ basic | ❌ |
| Rollback checkpoints | ✅ | ❌ | ❌ | ❌ | ✅ (git) | ❌ |
| Voice input/output | ✅ | ❌ | ❌ | ❌ | ⚠️ voice input | ❌ |
| Vision / image input | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ |
| Open source | ✅ MIT | ❌ | ❌ | ❌ | ✅ Apache | ⚠️ core MIT |
Capability comparisons are approximate and reflect the state of each tool as of early 2026. Check each project's docs for the latest.
License
AtlasMind is open-source under the MIT License.