LIFEHUBBER
Choose theme

AI Resources

OpenDuet

GitHub stars: 1.6K GitHub forks: 200 Declared license: MIT: MIT Last pushed October 28, 2026: Pushed 1d ago
Stats from GitHub

OpenDuet is an early-stage proactive-assistant template from CopilotKit, built around Gmail, Google Calendar, delegated tasks and an agent with its own browser and workspace.

It runs on OpenBot and adds a daily brief, task cards, channels and reusable skills. You host the application; CopilotKit Intelligence is required for its threads and memory. Use this as a first read, not a recommendation. Open the original project before trusting details like terms, limits, privacy, cost, setup, or safety.

What it is

A work assistant you run and configure

The project describes an assistant that reads connected mail and calendar sources, highlights what needs attention and can take on follow-up work. Its agent computer has a browser, files and a shell.

Why it stands out

A brief that can become a task

The morning brief groups mail by what needs you and what can wait, with reasons. Starting an item hands it to the assistant, while task cards show the work and any input or review it needs.

Availability

Early source-based setup

The repository declares MIT and provides a Docker-and-Bun quick start. It requires a model key and a CopilotKit Intelligence project. This is a developer setup, with no published OpenDuet image or hosted deployment guide yet.

Why it matters

What makes it useful

An inbox summary still leaves someone to choose the next job and carry it out. OpenDuet connects that choice to delegated work: a brief can lead into drafting or research, and the assistant keeps the task separate from the main chat. The documented workflow also identifies sources it could not check, so an unconnected calendar should not be mistaken for an empty day.

Notable points

What stands out

Task cards distinguish work in progress from work waiting for the person. Working, Ready for review and Needs input lead back to the task conversation rather than leaving every job in the main chat. Separately, an action set to ask pauses at the policy gateway with its proposed action and the rule that triggered it. The project documents Allow once, Always allow and Refuse choices; review the active rules rather than treating the presence of an approval screen as a guarantee about every action.

Before using

What to review

The project is in early development. Its quick start runs OpenDuet from source; the Docker and Helm deployment references under OpenBot describe the plain OpenBot image, not a published OpenDuet image.

Hosting the application does not keep every part local. Threads and memory live in CopilotKit Intelligence; model requests and connected services have their own data paths. The repository says the free Developer plan is enough to start, but model, hosting and other service costs are separate.

Review the Google account permissions, agent tool grants, approval rules and scheduled routines before connecting real work. Routines run as the person who configured them, and a container with a browser and shell can still act through the access it receives.

The telemetry documentation describes CopilotKit runtime usage events linked to the Intelligence project. It documents COPILOTKIT_TELEMETRY_DISABLED=true or DO_NOT_TRACK=1, followed by a server restart, to turn those events off.

The setup checklist separates Ready, Not checked and Needs setup. A configured address or a remembered setting alone is not evidence that a dependency is working.

Reader fit

Who may find it relevant

Developers exploring an assistant for email, calendar planning and follow-up tasks.

Teams that want to adapt a working assistant template and manage its hosting, model connection and app permissions.

Less suitable for someone seeking an offline assistant or a finished service requiring no infrastructure setup.

Editorial note

Why LifeHubber lists it

We list OpenDuet because its work-assistant behavior is packaged separately from the OpenBot platform beneath it. Its name, Bots, channels, model and skills live together in an editable tenant folder. For a developer building an assistant around their own work, the Gmail-and-calendar workflow provides a starting point to adapt, while the platform code remains separate for bringing in OpenBot fixes. That makes the project useful to explore as a customizable assistant, beyond trying a single fixed bot.

Source links

Source materials

Reader note

Before relying on this entry

LifeHubber lists entries to help readers inspect AI projects, not to endorse them or prove they are safe, suitable, accurate, maintained, or right for a specific use. We do not verify every entry in depth. Before relying on anything listed, review the original materials, terms, privacy practices, limits, and risks that matter for your situation.

Advertisements

Advertisements

For project maintainers

Listed here? You can use the badge.

If you maintain a project with a current LifeHubber listing, you may add the optional “Listed on LifeHubber AI Resources” badge to its README, docs, or website. No introduction or permission request is needed.

See what’s moving