Documentation Index
Fetch the complete documentation index at: https://bastani.mintlify.app/llms.txt
Use this file to discover all available pages before exploring further.
Why not markdown, a coding agent alone, or a general agent framework?
Markdown is great for guidance: conventions, commands, repo notes. Use Claude Code, OpenCode, or Copilot CLI directly for normal single-session coding.
Reach for Atomic at the point where the work needs branching, retries, parallel sessions, explicit state, human approval, sandboxed execution, or reliable handoff between stages. General agent frameworks can do some of this, but you often rebuild coding-agent basics yourself: file editing, terminal interaction, MCP setup, hooks, session handling, repo-specific context. Atomic starts from production coding agents and adds the workflow layer around them.
How does Atomic differ from Spec-Kit?
Spec-Kit is GitHub’s spec-driven development toolkit. Spec-Kit works well for greenfield projects with a single Copilot session. Atomic is built for the harder case — large existing codebases where you research first, with multi-session pipelines, isolated context windows, deterministic execution, and support for Claude Code, OpenCode, and Copilot CLI.
| Aspect | Spec-Kit | Atomic |
|---|
| Focus | Greenfield, spec-first | Large existing codebases — research-first or spec-first |
| Workflow definition | Shell scripts + markdown | TypeScript Workflow SDK with deterministic execution |
| Session management | Single agent session | Multi-session pipelines, sequential and parallel, in isolated context windows |
| Agent support | GitHub Copilot CLI | Claude Code + OpenCode + Copilot CLI |
| Sub-agents / skills | Single general agent / none | 12 specialized sub-agents + 57 built-in skills |
| Autonomous execution | Not available | Ralph — multi-hour autonomous plan/implement/review/debug loop |
How does Atomic differ from DeerFlow?
DeerFlow is ByteDance’s general-purpose agent runtime on LangGraph. DeerFlow is a general-purpose agent system with a web UI; Atomic is narrowly focused on coding workflows. Atomic runs on top of production coding agents (Claude Code, OpenCode, Copilot CLI) rather than reimplementing coding tools through a generic API — you get each agent’s native file editing, permissions, MCP integrations, and hooks out of the box.
| Aspect | DeerFlow | Atomic |
|---|
| Runtime | Python (LangGraph) | TypeScript (Bun) |
| Agent SDKs | OpenAI-compatible API | Claude Code + OpenCode + Copilot CLI native SDKs |
| Execution | DAG with conditional edges | Deterministic — strict step ordering, frozen definitions, controlled transcript passing |
| Sub-agents | Researcher / coder / reporter | 12 specialized sub-agents with scoped tools |
| Interface | Web UI (Streamlit) | Terminal chat with tmux session management |
| Autonomous | Not available | Ralph — bounded plan/implement/review/debug loop |
How does Atomic differ from Hermes Agent?
Hermes Agent is Nous Research’s general-purpose AI assistant with a self-improving learning loop. Hermes is a broad personal assistant that learns across sessions; Atomic is coding-specific workflow software for engineering teams. Atomic encodes your dev process as deterministic TypeScript workflows that run identically across teammates, machines, and CI, and inherits production-hardened tools from Claude Code, OpenCode, and Copilot CLI (giving you two independent security boundaries: devcontainer + agent permissions).
| Aspect | Hermes Agent | Atomic |
|---|
| Focus | Personal AI assistant (coding + messaging + smart home) | Coding-specific multi-session workflows for teams |
| Runtime | Python | TypeScript (Bun) |
| Agent SDKs | OpenAI-compatible adapter | Claude Code + OpenCode + Copilot CLI native SDKs |
| Execution | Single conversation loop with context compression | Multi-session pipelines, fresh context per session, controlled transcript passing |
| Skills | Auto-created (may drift) | 57 developer-authored, version-controlled skills |
| Reproducibility | Different paths each run | Frozen workflow definitions run identically across machines and CI |
| Security | Command approval + container backends | Devcontainer isolation + coding agent permission systems (two independent boundaries) |
