A local-first agent that can browse, inspect and edit files, run approved commands and keep session state on the user’s machine. Local inference through its llama.cpp-based runtime avoids an API subscription; optional cloud models and separately licensed model weights remain distinct.
Free MIT agent with local models; cloud use optional
The MIT application has a local llama.cpp model path without a required API subscription. Model weights are separate and keep their own terms; the checked Qwen2.5-Coder-7B-Instruct model is Apache-2.0, but check format compatibility and hardware before download. Cloud models can charge separately.
Atomic Agent · Frame from the official demo animation at 20 seconds; the displayed model and conversation are demo content, not an AltStack test.Product image: AtomicBot-ai and repository contributors · MIT License · Original image; layout scaled to fit. Frame from the official demo animation at 20 seconds; the displayed model and conversation are demo content, not an AltStack test. Full notice.Logo: AtomicBot-ai and repository contributors · MIT License · Original image; layout scaled to fit. Official project identity asset used only to identify this independent listing; no sponsorship or endorsement is implied. Full notice.
About Atomic Agent
A local-first agent that can browse, inspect and edit files, run approved commands and keep session state on the user’s machine. Local inference through its llama.cpp-based runtime avoids an API subscription; optional cloud models and separately licensed model weights remain distinct. A self-hosting option is recorded for this project. Review the official deployment instructions and maintenance requirements before running it yourself.
Release status
Official README and root license were source-checked on 3 November 2026. Local inference, hosting, API, model-weight and storage costs are described separately. AltStack did not execute this repository or its installers.
Follow the publisher’s current installation or quickstart guide for the edition and device you want to use. Tool-specific installation steps have not yet been checked by AltStack.
Open the official website or source repository below.
Check its supported devices and installation requirements, along with the pricing shown on this page.
Follow the provider’s instructions, then try a small non-sensitive sample before using important files.
Free MIT agent with local models; cloud use optional. The MIT application has a local llama.cpp model path without a required API subscription. Model weights are separate and keep their own terms; the checked Qwen2.5-Coder-7B-Instruct model is Apache-2.0, but check format compatibility and hardware before download. Cloud models can charge separately. Pricing evidence.
Devices / hardware
macOS, Linux, Windows. Check the linked provider instructions for versions and prerequisites.
No one-line command has been verified for Atomic Agent; follow its official installation instructions instead of guessing a Docker image or installer.
Connect an AI assistant to the tools and conversations where you already work.
The MIT-licensed assistant software is free to run on your own machine. Cloud models, connected services and hosted machines can require payment; subscriptions or API keys are supplied separately.
Use a desktop AI agent with persistent context and integrations across communication platforms.
The MIT-licensed agent can run locally without a software license fee. Model APIs, optional Nous Portal subscriptions, search or other services, and cloud machines may charge separately. A free agent is not free access to every model.
OpenCode can be used without a Go subscription. Optional Go is USD 10/month and Go Plus USD 40/month, with model-specific usage limits. Local models use your hardware; connected provider APIs and subscriptions bill separately. Promotional free models are limited-time offers.
Minimal terminal coding agent you can fully customize.
Official README supplies the installable Pi harness under MIT and instructs users to connect their own provider subscription or API key. The free harness does not include model usage or provider subscription fees.
Autonomous coding agents that complete real engineering tasks.
Official pricing lists MIT Agent Canvas free to self-host using own LLM key, Teams $499/month for up to 10 users with extra seats $50/month, and custom Enterprise. Provider tokens and managed cloud are separate.
Use an AI coding agent in an editor or terminal, with an optional local model.
The Apache-2.0 Cline VS Code extension and CLI have no software license charge. Use local Ollama models without a paid API key or per-request model charge, on your own hardware. Cline usage-billing, ClinePass subscriptions and external cloud model providers are optional paid routes with their own terms. The current repository says its JetBrains plugin is not open sourced. This review establishes the local client route, not unlimited hosted inference.
Allow Google Analytics to measure page visits and basic device information using cookies? Your searches and form entries are excluded. You can change this choice at any time. Privacy details.