Discord
0Computer Use Agent Platform - AComputerComputermacOSWindowsLinuxBrowserPhoneforAIAIAgentMCPAutomationEvalsRL
Build computer use agents that see screens, click buttons, type and run code
— with Cua
Loved by builders.
AK@_akhaliq
a new computer-use interface on Apple Silicon using local sandboxes by using ai-gradio
zingyzingy90210
#generalJust wanted everyone to know, the CUA team are brilliant. They look at every error you post. Honestly wasnt expecting a reply but they actually solved the issue I was having! I'd recommend non technical ppl here to create a github account and post any errors you have there as well as here. The team is constantly monitoring and fixing.
Tom Dörr@tom_doerr
VM for Agents. Just today, my agent setup broke my computer, preventing disk writing (which also means most programs won't start)
Adam Silverman (Hiring!) 🖇️@AtomSilverman
Hot take: The only world where browser use / computer use agents are successful at scale is if they are within a VM companies want their employees to work, not watch their agents do work while they relax
15 Trending AI Projects on GitHub: opcode, FastMCP, Dyad, RustGPT, Paper2Agent, Pepper, AG-UI,shimmy
Github Awesome
Watch video
Chris Barber@chrisbarber
AI Timelines: When will AI reach human-level in computer-use skills? I surveyed AI researchers and forecasters. I asked: by what quarter & year are you nearly certain (9-in-10 chance) that AI will reach human-level on the OSWorld computer-use benchmark? Surveyed: ,…
Tom Dörr@tom_doerr
What so my IDE now writes code and creates presentations with analysis results? New AI tools seem like April Fools' jokes
Garry Tan@garrytan
Use MCP to run computer use in a VM
Shubham | Agnost AI#laalshaitaan
#generalhey hey, excited to see how cua pans out! loved the product from whatever i could infer ❤️ all the best. a fellow oss contributor, would love to help in anyway possible, lmk if something pops up!
The Industry Reacts to GPT-5
Matthew Berman
Watch video
Adam Silverman (Hiring!) 🖇️@AtomSilverman
3/ The era of local Computer-Use AI Agents is here! introduces UI-TARS-1.5-7B-6bit, now running natively on Apple Silicon via MLX.
Solomon Hykes@solomonstre
OK we may have a path to running native mac builds in ! Thanks to the lume CLI by 😁 I'm thinking we could have the Dagger CLI natively run that, and expose it as a core primitive in the Dagger API. cc
Robert Scoble@Scobleizer
"Control full operating systems."
Unwind AI@unwind_ai_
2. Build Computer-Use AI Agents that control specific applications. Deploy a team of specialized lightweight agents in parallel, each focused on their own app, like "just control the iPhone Mirroring app."
TBPN@tbpn
We talked to , docker for computer-use agents. and are working on an open-source framework that enables AI agents to control full OS within high-performance, lightweight virtual containers. Currently, it delivers up to 90% native speed on Apple…
taruntarun0023
#generalThere's a million things that could be built on top of ollama and a million other things using cua, its the future! Embrace it, and don't forget to enjoy the process, results are just a by-product. These hackthons when done correctly have changed the entire course of one's life, I have personally seen this happen with some of own friends, the network effects, the sense of community and urgency that you experience for these next one week, will literally define your limits, and push you to the edge of safely operational capacity. If you need any help with resources / guidance, i am sure @James will be here to help everyone in any ways possible, he's one of the best in town. If you need anything from my end, to get a second opinion on somethings or even just someone to vent out to i am here, feel free to DM me. And you got this guys! If you are not living on the edge, then you are taking too much space. When you hold a hammer, everything looks like a nail. CUA team has given you the damn hammer go nail it
Yoko@stuffyokodraws
[New Post] CUA market and where things are. Below is what it takes to construct a computer use agent today, and some different approaches to get there We are still in the very early innings but and I are excited to hear what you build 🧵
Erik Thorelli@esthor
cua is 🔥 and their team is incredible. big congrats and for crushing it on computer use and doing it with OSS👏
Adam Silverman (Hiring!) 🖇️@AtomSilverman
21/ benchmarks Moondream3 against GTA-1. It delivers solid accuracy for open-weight agent models.
💥 \newline@newlinedotco
Benchmarking models is a breeze with cua. You can do it on popular datasets with just a single line of code using the HUD integration.
AK@_akhaliq
a new computer-use interface on Apple Silicon using local sandboxes by using ai-gradio
zingyzingy90210
#generalJust wanted everyone to know, the CUA team are brilliant. They look at every error you post. Honestly wasnt expecting a reply but they actually solved the issue I was having! I'd recommend non technical ppl here to create a github account and post any errors you have there as well as here. The team is constantly monitoring and fixing.
Tom Dörr@tom_doerr
VM for Agents. Just today, my agent setup broke my computer, preventing disk writing (which also means most programs won't start)
Adam Silverman (Hiring!) 🖇️@AtomSilverman
Hot take: The only world where browser use / computer use agents are successful at scale is if they are within a VM companies want their employees to work, not watch their agents do work while they relax
15 Trending AI Projects on GitHub: opcode, FastMCP, Dyad, RustGPT, Paper2Agent, Pepper, AG-UI,shimmy
Github Awesome
Watch video
Chris Barber@chrisbarber
AI Timelines: When will AI reach human-level in computer-use skills? I surveyed AI researchers and forecasters. I asked: by what quarter & year are you nearly certain (9-in-10 chance) that AI will reach human-level on the OSWorld computer-use benchmark? Surveyed: ,…
Tom Dörr@tom_doerr
What so my IDE now writes code and creates presentations with analysis results? New AI tools seem like April Fools' jokes
Garry Tan@garrytan
Use MCP to run computer use in a VM
Shubham | Agnost AI#laalshaitaan
#generalhey hey, excited to see how cua pans out! loved the product from whatever i could infer ❤️ all the best. a fellow oss contributor, would love to help in anyway possible, lmk if something pops up!
The Industry Reacts to GPT-5
Matthew Berman
Watch video
Adam Silverman (Hiring!) 🖇️@AtomSilverman
3/ The era of local Computer-Use AI Agents is here! introduces UI-TARS-1.5-7B-6bit, now running natively on Apple Silicon via MLX.
Solomon Hykes@solomonstre
OK we may have a path to running native mac builds in ! Thanks to the lume CLI by 😁 I'm thinking we could have the Dagger CLI natively run that, and expose it as a core primitive in the Dagger API. cc
Robert Scoble@Scobleizer
"Control full operating systems."
Unwind AI@unwind_ai_
2. Build Computer-Use AI Agents that control specific applications. Deploy a team of specialized lightweight agents in parallel, each focused on their own app, like "just control the iPhone Mirroring app."
TBPN@tbpn
We talked to , docker for computer-use agents. and are working on an open-source framework that enables AI agents to control full OS within high-performance, lightweight virtual containers. Currently, it delivers up to 90% native speed on Apple…
taruntarun0023
#generalThere's a million things that could be built on top of ollama and a million other things using cua, its the future! Embrace it, and don't forget to enjoy the process, results are just a by-product. These hackthons when done correctly have changed the entire course of one's life, I have personally seen this happen with some of own friends, the network effects, the sense of community and urgency that you experience for these next one week, will literally define your limits, and push you to the edge of safely operational capacity. If you need any help with resources / guidance, i am sure @James will be here to help everyone in any ways possible, he's one of the best in town. If you need anything from my end, to get a second opinion on somethings or even just someone to vent out to i am here, feel free to DM me. And you got this guys! If you are not living on the edge, then you are taking too much space. When you hold a hammer, everything looks like a nail. CUA team has given you the damn hammer go nail it
Yoko@stuffyokodraws
[New Post] CUA market and where things are. Below is what it takes to construct a computer use agent today, and some different approaches to get there We are still in the very early innings but and I are excited to hear what you build 🧵
Erik Thorelli@esthor
cua is 🔥 and their team is incredible. big congrats and for crushing it on computer use and doing it with OSS👏
Adam Silverman (Hiring!) 🖇️@AtomSilverman
21/ benchmarks Moondream3 against GTA-1. It delivers solid accuracy for open-weight agent models.
💥 \newline@newlinedotco
Benchmarking models is a breeze with cua. You can do it on popular datasets with just a single line of code using the HUD integration.
AK@_akhaliq
a new computer-use interface on Apple Silicon using local sandboxes by using ai-gradio
zingyzingy90210
#generalJust wanted everyone to know, the CUA team are brilliant. They look at every error you post. Honestly wasnt expecting a reply but they actually solved the issue I was having! I'd recommend non technical ppl here to create a github account and post any errors you have there as well as here. The team is constantly monitoring and fixing.
Tom Dörr@tom_doerr
VM for Agents. Just today, my agent setup broke my computer, preventing disk writing (which also means most programs won't start)
Adam Silverman (Hiring!) 🖇️@AtomSilverman
Hot take: The only world where browser use / computer use agents are successful at scale is if they are within a VM companies want their employees to work, not watch their agents do work while they relax
15 Trending AI Projects on GitHub: opcode, FastMCP, Dyad, RustGPT, Paper2Agent, Pepper, AG-UI,shimmy
Github Awesome
Watch video
Chris Barber@chrisbarber
AI Timelines: When will AI reach human-level in computer-use skills? I surveyed AI researchers and forecasters. I asked: by what quarter & year are you nearly certain (9-in-10 chance) that AI will reach human-level on the OSWorld computer-use benchmark? Surveyed: ,…
Tom Dörr@tom_doerr
What so my IDE now writes code and creates presentations with analysis results? New AI tools seem like April Fools' jokes
Garry Tan@garrytan
Use MCP to run computer use in a VM
Shubham | Agnost AI#laalshaitaan
#generalhey hey, excited to see how cua pans out! loved the product from whatever i could infer ❤️ all the best. a fellow oss contributor, would love to help in anyway possible, lmk if something pops up!
The Industry Reacts to GPT-5
Matthew Berman
Watch video
Adam Silverman (Hiring!) 🖇️@AtomSilverman
3/ The era of local Computer-Use AI Agents is here! introduces UI-TARS-1.5-7B-6bit, now running natively on Apple Silicon via MLX.
Solomon Hykes@solomonstre
OK we may have a path to running native mac builds in ! Thanks to the lume CLI by 😁 I'm thinking we could have the Dagger CLI natively run that, and expose it as a core primitive in the Dagger API. cc
Robert Scoble@Scobleizer
"Control full operating systems."
Unwind AI@unwind_ai_
2. Build Computer-Use AI Agents that control specific applications. Deploy a team of specialized lightweight agents in parallel, each focused on their own app, like "just control the iPhone Mirroring app."
TBPN@tbpn
We talked to , docker for computer-use agents. and are working on an open-source framework that enables AI agents to control full OS within high-performance, lightweight virtual containers. Currently, it delivers up to 90% native speed on Apple…
taruntarun0023
#generalThere's a million things that could be built on top of ollama and a million other things using cua, its the future! Embrace it, and don't forget to enjoy the process, results are just a by-product. These hackthons when done correctly have changed the entire course of one's life, I have personally seen this happen with some of own friends, the network effects, the sense of community and urgency that you experience for these next one week, will literally define your limits, and push you to the edge of safely operational capacity. If you need any help with resources / guidance, i am sure @James will be here to help everyone in any ways possible, he's one of the best in town. If you need anything from my end, to get a second opinion on somethings or even just someone to vent out to i am here, feel free to DM me. And you got this guys! If you are not living on the edge, then you are taking too much space. When you hold a hammer, everything looks like a nail. CUA team has given you the damn hammer go nail it
Yoko@stuffyokodraws
[New Post] CUA market and where things are. Below is what it takes to construct a computer use agent today, and some different approaches to get there We are still in the very early innings but and I are excited to hear what you build 🧵
Erik Thorelli@esthor
cua is 🔥 and their team is incredible. big congrats and for crushing it on computer use and doing it with OSS👏
Adam Silverman (Hiring!) 🖇️@AtomSilverman
21/ benchmarks Moondream3 against GTA-1. It delivers solid accuracy for open-weight agent models.
💥 \newline@newlinedotco
Benchmarking models is a breeze with cua. You can do it on popular datasets with just a single line of code using the HUD integration.
AK@_akhaliq
a new computer-use interface on Apple Silicon using local sandboxes by using ai-gradio
zingyzingy90210
#generalJust wanted everyone to know, the CUA team are brilliant. They look at every error you post. Honestly wasnt expecting a reply but they actually solved the issue I was having! I'd recommend non technical ppl here to create a github account and post any errors you have there as well as here. The team is constantly monitoring and fixing.
Tom Dörr@tom_doerr
VM for Agents. Just today, my agent setup broke my computer, preventing disk writing (which also means most programs won't start)
Adam Silverman (Hiring!) 🖇️@AtomSilverman
Hot take: The only world where browser use / computer use agents are successful at scale is if they are within a VM companies want their employees to work, not watch their agents do work while they relax
15 Trending AI Projects on GitHub: opcode, FastMCP, Dyad, RustGPT, Paper2Agent, Pepper, AG-UI,shimmy
Github Awesome
Watch video
Chris Barber@chrisbarber
AI Timelines: When will AI reach human-level in computer-use skills? I surveyed AI researchers and forecasters. I asked: by what quarter & year are you nearly certain (9-in-10 chance) that AI will reach human-level on the OSWorld computer-use benchmark? Surveyed: ,…
Tom Dörr@tom_doerr
What so my IDE now writes code and creates presentations with analysis results? New AI tools seem like April Fools' jokes
Garry Tan@garrytan
Use MCP to run computer use in a VM
Shubham | Agnost AI#laalshaitaan
#generalhey hey, excited to see how cua pans out! loved the product from whatever i could infer ❤️ all the best. a fellow oss contributor, would love to help in anyway possible, lmk if something pops up!
The Industry Reacts to GPT-5
Matthew Berman
Watch video
Adam Silverman (Hiring!) 🖇️@AtomSilverman
3/ The era of local Computer-Use AI Agents is here! introduces UI-TARS-1.5-7B-6bit, now running natively on Apple Silicon via MLX.
Solomon Hykes@solomonstre
OK we may have a path to running native mac builds in ! Thanks to the lume CLI by 😁 I'm thinking we could have the Dagger CLI natively run that, and expose it as a core primitive in the Dagger API. cc
Robert Scoble@Scobleizer
"Control full operating systems."
Unwind AI@unwind_ai_
2. Build Computer-Use AI Agents that control specific applications. Deploy a team of specialized lightweight agents in parallel, each focused on their own app, like "just control the iPhone Mirroring app."
TBPN@tbpn
We talked to , docker for computer-use agents. and are working on an open-source framework that enables AI agents to control full OS within high-performance, lightweight virtual containers. Currently, it delivers up to 90% native speed on Apple…
taruntarun0023
#generalThere's a million things that could be built on top of ollama and a million other things using cua, its the future! Embrace it, and don't forget to enjoy the process, results are just a by-product. These hackthons when done correctly have changed the entire course of one's life, I have personally seen this happen with some of own friends, the network effects, the sense of community and urgency that you experience for these next one week, will literally define your limits, and push you to the edge of safely operational capacity. If you need any help with resources / guidance, i am sure @James will be here to help everyone in any ways possible, he's one of the best in town. If you need anything from my end, to get a second opinion on somethings or even just someone to vent out to i am here, feel free to DM me. And you got this guys! If you are not living on the edge, then you are taking too much space. When you hold a hammer, everything looks like a nail. CUA team has given you the damn hammer go nail it
Yoko@stuffyokodraws
[New Post] CUA market and where things are. Below is what it takes to construct a computer use agent today, and some different approaches to get there We are still in the very early innings but and I are excited to hear what you build 🧵
Erik Thorelli@esthor
cua is 🔥 and their team is incredible. big congrats and for crushing it on computer use and doing it with OSS👏
Integrate
Computer Use Stack
One unified interface for computer use models. Execute powerful workflows with structured outputs, multi-turn conversations, and custom tools.
Choose your integration
# Learn more: cua.ai/docs
import os
from agent import ComputerAgent
from computer import Computer
# Set your API key from cua.ai/dashboard
os.environ["CUA_API_KEY"] = "sk_..."
# Initialize cloud computer (create from dashboard or CLI)
computer = Computer(
os_type="linux",
provider_type="cloud",
name="your-sandbox-name"
)
# Initialize agent with model and tools
agent = ComputerAgent(
model="cua/anthropic/claude-sonnet-4.5",
tools=[computer]
)
# Run agent with task
messages = [{"role": "user", "content": "Take a screenshot"}]
async for result in agent.run(messages):
print(result)Ship and Scale
Ship and Scale Fast
Cloud-powered sandboxes for your computer use agents. Simple API integration, unlimited scale.
One-line deploy
Create sandboxes, run tasks, and manage everything from your terminal
cua-cli
$ cua sb create --os linux --size small
Scale Effortlessly
Run as many sandboxes as you need. No infrastructure management, no resource limits
Active
Active
Active
Active
Cross-platform
Choose between Linux, Windows, and macOS sandboxes based on your automation needs
Preview
Cloud VLM Inference
Access 100+ vision-language models from top providers with one API key
Pay-as-you-go
Only pay for what you use. Credit-based billing with no upfront costs or monthly minimums
Smart auto-routing
Automatically route requests to the best model for your task, balancing performance and cost
Open source
Agent SDK, Computer SDK, macOS virtualization, and Docker images. All open source on GitHub
Data & Benchmarks
Data, Train and Evaluate
Generate large-scale UI datasets, capture agent trajectories, and run standardized benchmarks. Everything you need to train and evaluate computer-use agents.
UI dataset generation
Generate diverse UI screenshots with ground-truth bounding boxes, labels, and metadata for ML training
bbox
label
Trajectory recording
Capture multi-step agent interactions with full state tracking. Replay, intervene, and create training data
reset
t=0click
t=1type
t=2scroll
t=3done
t=4Benchmark evals
Evaluate agents on safety, click accuracy, and multi-step tasks. From basic interactions to complex workflows
Safety
Click Accuracy
Multi-step Tasks
Form Filling
HuggingFace export
Export datasets as Arrow/Parquet for ML training. Push directly to HuggingFace Hub with one command
.arrow
.parquet
.jsonl
HuggingFace Hub
Multi-OS environments
Generate datasets across macOS, Windows, and Linux with authentic OS chrome, icons, and layouts
Reproducible tasks
Seeded randomization and deterministic environments. Replay from any checkpoint for debugging
Trajectory viewer
Visual replay of agent runs with step-by-step navigation. Freeze and intervene at any point
Pricing that scales with you
Pay only for what you use - credits for cloud compute and VLM inference
I am looking to...
Free
Free
10 credits to start
Try before you commit - perfect for testing and evaluation
- Linux Small sandbox only
- Claude Sonnet 4.5 Haiku only
- Community Discord support
- MIT License open source
Most Popular
Pro
$10+
Flexible credits for cloud compute and Computer-Use VLMs, with managed cloud environments and inference.
- Cloud Linux and Windows Environments
- CUA LLM Inference Provider
- Cuazar Playground UI
- Priority support via Slack
$
100 credits per dollar
Enterprise
Custom monthly
Scalable cloud containers tailored for large teams and organizations.
- Everything in Pro
- 24/7 support
- HIPAA, SOC Type 1/2 Reports
What do credits buy?
Credits are our unified currency for both compute time and AI inference
How many credits do you need per month?
10kcredits/month
Approximately $100/month
Compute Time
~2,000 hours
on Medium Linux instance
VLM Inference
~7.7M tokens
on Claude Sonnet
100
1k
5k
10k
25k
50k
100k
Sandbox Compute
Per hour for cloud environments
Linux Small5 credits/hour
Linux Medium9 credits/hour
Linux Large24 credits/hour
Windows Small8 credits/hour
Windows Medium15 credits/hour
Windows Large31 credits/hour
Billed per minute. Sessions include full desktop access with persistent storage.
VLM Inference
Per million tokens processed
Gemini 3 Flash Preview~254 credits/1M
Claude Haiku 4.5~435 credits/1M
Claude Sonnet 4.5~1305 credits/1M
Gemini 3 Pro Preview~1595 credits/1M
Claude Opus 4.5~2175 credits/1M
Fara-7B~2175 credits/1M
ByteDance UI-TARS-2~2610 credits/1M
Qwen3 VL 30B~8050 credits/1M
Input and output tokens combined. Actual costs vary by model and usage.
FAQ
Frequently Asked Questions
Everything you need to know about CUA and computer-use agents
1 minute to your first agent
No DevOps. No infrastructure. Just ship.
Agent Runs to Date
No credit card required!