Run useful routines
Schedule research, monitoring, reports, and other repeatable jobs. Review every run and its results from the workbench.
omnideck
Own your
AI workbench.
Put AI agents to work on research, projects, and recurring tasks from one desktop app. The workbench runs on your computer, and you choose the models and tools each agent can use.
Download the free public beta for macOS, Windows 11, or Linux. No omnideck account required.
- Open source
- Runs locally
- Cloud + local models
Your model provider shouldn’t own your workbench.
Proprietary AI platforms often bundle the model, workspace, automation, and your history into one closed service. omnideck keeps the workbench open and on your machine. You choose what agents can access, whether each job uses a local or hosted model, and when work leaves your computer.
Research, create, and automate from one workbench.
Put local agents to work across the web and your connected services. Keep the results as artifacts or schedule the work to run again.
-
01 —
Research and act on the web
Have an agent compare sources, collect information, and move through browser-based work while you watch the live browser or take over when needed.
Live browser · Hugging Face, Ollama, and vLLM -
02 —
Create work you can use
Turn the research into a report, spreadsheet, document, image, or working app. Each result stays in the workbench as an artifact you can inspect, refine, and download.
Artifact preview · Model choice brief
Complete work across your services
Connect email, calendars, cloud storage, contacts, and APIs so omnideck can take action across the services you already use—not just tell you what to do next.
The workbench runs on your machine.
You choose the files, services, and models an agent can use. omnideck keeps those boundaries local, visible, and under your control.
Your machine is the workspace
omnideck and its agent runtime run locally, so your workbench is not tied to a proprietary AI platform.
Agents stay inside the boundary
Agent work runs in an isolated container with access limited to the shared files and system commands you allow.
Raw keys stay out of reach
API keys and OAuth tokens live in an encrypted local vault managed outside the agent process.
You decide what can leave
Choose when a job uses a hosted model or connected service. omnideck itself adds no telemetry or tracking.
Set it up around the work you do.
Create reusable agents for different kinds of work, then choose the model and tools each one should use.
Build the workbench your work needs.
Add new actions, repeatable ways of working, and focused interfaces without waiting for them to become built-in features.
Connect new actions
Connect APIs, local commands, and services so agents can carry work through to completion.
Reuse the way you work
Package instructions and resources that agents can apply whenever the same kind of work returns.
Give workflows an interface
Build focused interfaces for repeated workflows and run them directly inside the workbench.
Start in three steps.
Install the app, choose a model, and give an agent its first job.
Install omnideck
Choose the installer for your computer and follow setup into your local workbench.
Choose a model
Connect a hosted provider or use Ollama for models running on your machine.
Give an agent a job
Open the workbench and give an agent a job—research a question, create something useful, or set up work to run again.
install omnideck
Your workbench. Your models.
Your workbench. Your models.
Your machine.
omnideck is open source and free to run. Install the app and put your local workbench to work.