Atomic Agent is an open-source terminal-based AI coding agent developed by AtomicBot-ai. It provides an automated development loop designed to analyze local codebases, plan complex refactors, and execute terminal commands and file edits directly from developer workstations. The tool operates primarily via an interactive Terminal User Interface (TUI) that generates structured, actionable implementation plans. For model backends, Atomic Agent supports both external API endpoints and a fully managed local inference pipeline powered by embedded llama-server instances running GGUF models (such as Qwen 3.5 variants). A core differentiator for Atomic Agent is its native local model integration. It uses custom GBNF grammars to enforce strict JSON syntax for tool calling during sampling and correctly handles thinking/reasoning preludes from models with <think> templates, avoiding token-generation stalls common in local tool execution. Atomic Agent is open-source software distributed on GitHub. When paired with local GGUF weights, it runs without third-party API keys or recurring per-token subscription costs.
Interactive TUI Planning — Terminal UI presenting step-by-step actionable plans and real-time execution event streams. - Managed Local LLM Inference — Built-in lifecycle management for running local GGUF models through an embedded llama-server process. - GBNF Grammar Constraints — Enforces valid tool-call JSON schemas at the sampler level using GBNF grammar files. - Reasoning Model Support — Native handling of reasoning templates (including Qwen 3.5 thinking profiles) prior to tool invocation. - External and Managed Modes — Flexibly switch between fully managed local models and external OpenAI-compatible model endpoints. - Automated Tool Orchestration — Inspects files, applies diffs, and runs terminal commands within project workspaces.
Use cases
Use Case: Local and Air-Gapped Development Developers working under strict privacy or air-gapped constraints can run Atomic Agent completely offline using local GGUF models via llama-server without leaking code to external APIs.
Use Case: Grammar-Constrained Tool Invocation Teams running smaller open-weight models (such as 4B or 35B parameter weights) can rely on Atomic Agent's GBNF grammar masks to guarantee that tool calls strictly conform to valid JSON schemas without formatting errors.
Use Case: Interactive Code Refactoring Engineers can launch the TUI to generate an actionable plan for multi-file refactoring tasks, inspect diffs interactively, and approve step-by-step test execution.
Who it is for
Atomic Agent is designed for developers who prefer terminal-native workflows and want fine-grained control over model inference, particularly those utilizing open-weight local LLMs. It is less suitable for users looking for plug-and-play browser interfaces or non-technical AI assistants.
Install guide
Installation Install Atomic Agent from source or globally via npm (standard Node.js distribution method for the repository): bash # Install via git clone and build git clone https://github.com/AtomicBot-ai/atomic-agent.git cd atomic-agent npm install npm run build
First Run Start the agent from your project directory to configure model backend options (managed local llama-server or external endpoint): bash npx atomic-agent
Example Session Launch the TUI and provide a prompt to start an agentic coding workflow: bash # Inside the Atomic Agent interactive TUI prompt: > Analyze src/index.ts, refactor error handling to use safe extractors, and run npm test to verify changes.
ActPlane — ActPlane is an open-source agent execution and policy enforcement harness developed by the Eunomia-bpf organization. It functions as a security…
hol-guard — hol-guard is a security scanner and agent guardrail harness developed by Hashgraph Online. Built to safeguard agentic runtime environments and…
Skillreaper — Skillreaper is an open-source skill lifecycle management and auditing tool for AI agent ecosystems, developed by thousandflowers. It analyzes agent…
Agent Island — Agent Island is an open-source agent orchestration harness created by developer tristan666666 on GitHub. Designed as a framework for running…
Stratless — Stratless is an open-source agent harness and execution orchestrator developed by the Stratless AI team. It provides a modular infrastructure…
Google AX — Google AX is an open-source agent harness and execution orchestrator developed by Google. It provides runtime infrastructure and execution sandboxing…