Agent – A Local Computer-Use Operator for macOS
Hey HN! We've just open-sourced Agent, our framework for running computer-use workflows across multiple apps in isolated macOS/Linux sandboxes. After launching Computer a few weeks ago, we realized many of you wanted to run complex workflows that span multiple applications. Agent builds on Computer to make this possible. It works with local Ollama models (if you're privacy-minded) or cloud providers like OpenAI, Anthropic, and others. Why we built this: We kept hitting the same problems when building multi-app AI agents - they'd break in unpredictable ways, work inconsistently across…
What it does
In the maker’s words, at launch
Hey HN! We've just open-sourced Agent, our framework for running computer-use workflows across multiple apps in isolated macOS/Linux sandboxes. After launching Computer a few weeks ago, we realized many of you wanted to run complex workflows that span multiple applications. Agent builds on Computer to make this possible. It works with local Ollama models (if you're privacy-minded) or cloud providers like OpenAI, Anthropic, and others. Why we built this: We kept hitting the same problems when building multi-app AI agents - they'd break in unpredictable ways, work inconsistently across environments, or just fail with complex workflows. So we built Agent to solve these headaches: • It handles complex workflows across multiple apps without falling apart • You can use your preferred model (local or cloud) - we're not locking you into one provider • You can swap between different agent loop implementations depending on what you're building • You get clean, structured responses that work well with other tools The code is pretty straightforward: async with Computer() as macos_computer: agent = ComputerAgent( computer=macos_computer, loop=AgentLoop.OPENAI, model=LLM(provider=LLMProvider.OPENAI) ) tasks = [ "Look for a repository named trycua/cua on GitHub.", "Check the open issues, open the most recent one and read it.", "Clone the repository if it doesn't exist yet." ] for i, task in enumerate(tasks): print(f"\nTask {i+1}/{len(tasks)}: {task}") async for result in agent.run(task): print(result) print(f"\nFinished task {i+1}!") Some cool things you can do with it: • Mix and match agent loops - OpenAI for some tasks, Claude for others, or try our experimental OmniParser • Run it with various models - works great with OpenAI's computer_use_preview, but also with Claude and others • Get detailed logs of what your agent is thinking/doing (super helpful for debugging) • All the sandboxing from Computer means your main system stays protected Getting started is easy: pip install "cua-agent[all]" # Or if you only need specific providers: pip install "cua-agent[openai]" # Just OpenAI pip install "cua-agent[anthropic]" # Just Anthropic pip install "cua-agent[omni]" # Our experimental OmniParser We've been dogfooding this internally for weeks now, and it's been a game-changer for automating our workflows. Grab the code at https://github.com/trycua/cua, or join us on Discord if you have questions: https://discord.com/invite/mVnXXpdE85 What are you planning to automate? I'm especially curious about complex workflows that span multiple apps - those are the hardest to get right and where we've seen Agent really shine. Would love to hear your thoughts!
Pricing, as stated on its site
Free — Agent is an open-source framework available on GitHub with no pricing mentioned. The product itself appears to be free/open-source software. The pricing page shown is for GitHub's hosting platform, no
checked September 2026 · prices change
Does the same job
all alternatives →- AUAgency – Unifying human, AI, and other computing systems, in Python2023 · github.com · ▲106
Hello hackernews! I'm excited to share a new open source python library I just released for creating AI agent-integrated systems. The name is `agency`. It differs from other agent libraries, most importantly in that it's intended to address a distinct part of the overall problem, that of agent integration. It is not an agent toolchain like LangChain and others. `agency` is a framework intended for safely integrating agents with computing systems and humans in a way that all parties can easily understand and communicate with each other. I've spent a lot of time on the readme which contains a…
More ai this month
the category →
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.
AI · 16d ago · simedw.com
Astute▲585Automate your B2B brand going viral, with new media creators
AI · 18d ago · company-app.joinastute.com
Hey HN, Henry from Cactus here! We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2. The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400-1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges…
AI · 26d ago · cactuscompute.com
Launched alongside, March 2025
the whole month →
Mimic Human Research & Save Findings in AI Knowledge Base
AI · 2025 · sider.ai
