Activity stays live.
Return to work in progress, approvals, and deliveries.
Open-source desktop app
AI agent for work, on your computer
An AI that does more than answer: it plans, researches, uses your files and tools, and shows you every step. Local and private by default.
Local
Cloud
No mandatory accountMeaningful local value from the start
Local and private by default.Cloud when you choose it.
Open-source desktop appBuilt for work on your computer
Open exitsKeep portable data and deliverables
One request, visible all the way through
Plan. Execute. Verify. Deliver.
Follow the work through research, code, verification, and the decision loqy needs before anything consequential happens.
Follow the work. Track what matters.
See what your AI agent is doing, what’s waiting, and what needs you.
Track what matters.
Create a task or link work already in progress.
Your tasks, two views.
Use the list for detail or the board for progress.
Approvals, questions, active work, and recent deliveries, organized by what needs you now.
Local and private by default.
Choose the engine. Keep the same agent.
Run locally
Keep AI model interactions local.
Built-in local AI models
Choose from several integrated AI models matched to your system.
Built-in models
Ollama
Use local models from your Ollama service.
LM Studio
Connect LM Studio and load the local model you need.
Use cloud AI models
Cloud when you choose it.
BYOK available
Use your own API key with a supported provider.
Supported BYOK providers
Models available through Umans
One AI agent, many kinds of work
Research, create, and act across the whole job.
Review and synthesis
Turn contracts into a decision-ready review.
Find obligations and differences, with their sources.
Contract comparison
Two differences, with source citations.
Research and briefing
Research across files, data, and the live web.
Cross-check the web against your documents and trace the sources.
Evidence-backed brief
Citations stay attached
Code and artifacts
Write and run code to create checked outputs.
Turn your data into verified spreadsheets, presentations, or interactive results.
Actions and operations
Take controlled action with authorized tools.
Connect your tools and define the actions they can take.
Waiting for your approval
Export reviewed actions
Only the approved action set.
Planning folderapproved-actions.csv
Trust at the point of consequence
Privacy should be structural, not a setting.
You choose local or cloud.
Your AI model never changes silently.
Tools work in isolation.
Browser and code run in a dedicated microVM.
You approve sensitive actions.
Review the exact action and destination before approval.
On your device
Pipeline export.csv
Targets.csvIsolated work
Public web and analysis code
Pipeline reviewExact decision
Export approved actions to the Planning folder?
Waiting for youApproved destination
Planning folder
Inactive until approved
Sources, code, decisions, and output revisions stay attached to the work trail.
Data and deliverables remain exportable.
macOS
Apple Silicon, 16 GB or more
Windows
Linux
Web
Reviews, approvals, and work follow-up