What if your AI assistant could actually do the work like dots from openai? Meet OpenDots

Most AI assistants follow a simple pattern: you ask a question, and the AI gives you an answer.

But what if your AI assistant could do more than answer questions? What if it could research a topic, open websites, work with files, prepare documents, connect to your applications, and keep working on tasks while you focus on something else?

That is the idea behind OpenDots, an open-source AI workspace developed by CopilotKit.

OpenDots provides a starting point for building your own persistent AI coworkers, called Dots. Each Dot can have a specific role, instructions, tools, and permissions, allowing you to create specialized assistants for different types of work.

What exactly is OpenDots?

OpenDots is an open-source software template that you can run on your own infrastructure and customize for your needs. It is not a ready-made hosted AI product that works automatically after installation.

Instead, it provides a foundation for building an AI workspace where assistants can interact with users, use permitted tools, work with a computer environment, and save useful results.

For example, you could create different Dots for researching competitors, drafting marketing content, investigating technical issues, or preparing business documents.

The important difference is that OpenDots is designed around completing tasks, not just generating text.

1. AI coworkers with specialized roles

OpenDots allows you to configure individual AI assistants with their own names, roles, instructions, and permitted tools.

For example:

• Research assistant: Investigates a topic, gathers information, and prepares research notes.
• Writing assistant: Creates articles, drafts documents, and revises existing content.
• Business assistant: Works with connected services to help with email, calendar, and other business workflows, subject to configured permissions.
• Technical assistant: Uses its available tools to inspect websites, work with files, or assist with development tasks.

Each Dot can operate with its own conversation and configured permissions. You can inspect its work and decide which tools it is allowed to use.

This makes it possible to build an AI workspace tailored to your specific business processes.

2. AI assistants that can use a computer

One of OpenDots' interesting features is its support for dedicated computer environments for individual Dots.

Through its OpenBot-based computer services, a configured Dot can interact with an isolated browser and workspace. Depending on the setup and permissions, it can browse websites, work with files, execute shell commands, and display activity and results inside the application.

Imagine asking an assistant to research a competitor's website, summarize the important information, save the findings to a file, and verify that the file was created successfully.

Instead of receiving only instructions on how to perform these steps, the assistant can use its configured computer tools to carry out the workflow.

These capabilities require the relevant computer services to be installed and connected. They are not automatically available in an unconfigured installation.

3. Spaces: A workspace for documents and knowledge

OpenDots includes a feature called Spaces, which acts as a home for working documents.

You can create and organize pages, browse them in a searchable library, and arrange documents into nested pages. The editor supports formatting, slash commands, undo and redo, and Markdown source editing.

You can also save a conversation as a page or ask an AI specialist to create and revise content.

For example, a marketing team could organize its work into a Space containing:

• Marketing strategies
• Competitor research
• Campaign briefs
• Content calendars
• Draft articles

The assistant can then help you work with relevant saved pages, depending on the context and permissions configured for the conversation.

4. Human approval before important actions

Giving AI access to tools is useful, but allowing it to make every decision automatically can be risky.

OpenDots supports human-in-the-loop approval for important workflows.

When a Dot prepares a draft that needs review, the application can display an approval card with options to approve and save or decline the draft.

Similarly, tool integrations can pause before executing actions that require approval.

This creates a workflow in which the AI prepares the work, you review it, and the application proceeds only after the required approval.

For businesses, this approach is especially useful when dealing with customer communications, documents, or changes made through connected services.

Important: Approval behavior depends on the tool and integration being used. Review the permissions and confirmation requirements of each workflow before deploying it.

5. Connect AI coworkers to external applications

OpenDots supports integrations that extend what an assistant can do.

It includes support for Model Context Protocol (MCP) servers, which allow compatible tools and services to be exposed to AI assistants. It also supports connected applications through Composio.

Depending on your configuration, you can connect services such as email, calendars, GitHub, or your own applications.

For example, an assistant could help prepare an email, gather information from a connected service, or organize a task using an available integration.

You can enable or disable tools for individual Dots and configure which connected-app toolkits they can access.

Connected applications require setup, authentication, and appropriate permissions. Adding credentials alone does not automatically grant every assistant access to every account.

6. Work through chat, voice calls, and Slack

OpenDots is designed to support more than a conventional text-based chatbot.

Chat

You can interact with a Dot through a continuous conversation. The interface can display work updates, source links, tool activity, and results.

Voice calls

The project includes support for real-time voice conversations, with separate compute agents able to handle longer-running work. Voice requires additional provider configuration.

Slack

OpenDots includes an integration path for bringing specialist Dots into Slack through the Channels SDK and a managed connection.

This can help teams access AI assistance in the communication environment they already use. However, the Slack integration requires additional configuration, and connected-service verification may be necessary.

7. Background tasks and persistent work

OpenDots includes infrastructure for background work and scheduled tasks.

This allows configured tasks to continue processing outside the immediate interaction, with pause and retry controls available for supported workflows.

For example, you might configure an assistant to carry out a recurring research task or prepare a document in the background.

However, scheduled instructions are not the same as a complete autonomous goal-management system. The project documentation identifies limitations around event triggers, task responsibility, and automatic delegation between multiple Dots.

If you want an assistant to react to a specific business event, such as a new lead arriving in your CRM, you may need to build additional triggers and integrations.

8. How does OpenDots work technically?

OpenDots combines a web application with an agent runtime, model providers, tool integrations, and optional computer and communication services.

At a simplified level, the workflow looks like this:

1. You submit a task: You ask a Dot to research, write, or perform an operation.
2. The AI processes the request: The configured model interprets the task and determines which available tools may help.
3. The runtime coordinates the work: CopilotKit and AG-UI support communication between the interface and the agent runtime.
4. The assistant uses permitted tools: Depending on the setup, it can access a browser, workspace files, connected applications, or other configured services.
5. You review the result: The assistant presents its output, and approval is requested when required by the workflow.
6. The result is saved or returned: Documents and workspace information can be stored, while conversation data is persisted through the configured Intelligence deployment.

The project uses technologies including React, CopilotKit, AG-UI, TanStack AI, and SQLite for workspace metadata and pages. Its conversation persistence relies on the configured CopilotKit Intelligence deployment.

9. Can you run OpenDots locally or host it yourself?

Yes. OpenDots is an open-source template that you can clone and customize. Its repository is available on GitHub:

https://github.com/CopilotKit/OpenDots

The documented initial setup uses Node.js 24 and npm.

A simplified setup is:

git clone https://github.com/CopilotKit/OpenDots.git
cd OpenDots
npm ci
cp .env.example .env
npm run dev

You can then open the local application at http://127.0.0.1:5173.

This initial setup lets you explore the workspace and configure Dots. To enable conversations, you must configure a compatible model and CopilotKit Intelligence.

Additional setup is required for computer services, voice, Slack, and connected applications. The repository also documents a local Docker evaluation option for Intelligence, while a licensed self-hosted deployment is available through CopilotKit.

Running the application locally does not automatically mean that every AI model, conversation service, or connected integration runs locally. Check each provider's requirements before choosing your deployment architecture.

10. Is OpenDots completely free?

The OpenDots source code is available under the MIT license, which makes the template available for use and customization under that license.

However, the overall cost depends on how you deploy and operate it.

You may need to account for:

• Server or computer infrastructure
• AI model usage
• CopilotKit Intelligence or an eligible self-hosted deployment
• Voice and communication providers
• Connected application services
• Maintenance and deployment costs

Therefore, open source does not necessarily mean that every service required by OpenDots is free.

If you are evaluating it for a business, calculate the infrastructure and provider costs before using it in production.

11. OpenDots limitations to understand

OpenDots is promising, but it is important to understand what it does not provide out of the box.

• It is a template, not a fully managed product. You are responsible for configuring and maintaining your deployment.
• Integrations need setup. Browser tools, computer services, voice, Slack, and connected applications have their own configuration requirements.
• Security requires careful configuration. Connected tools may act on real accounts, so access should be restricted and tested.
• It is still evolving. The repository documents areas where live integration verification or additional development is still required.
• It is not a complete multi-user enterprise platform. Shared editing, invitations, and some collaboration features are not included in the current template.
• Multiple specialist agents do not automatically mean full agent collaboration. Multi-Dot group conversations and automatic delegation remain additional development areas.

These limitations do not make the project less useful. They simply mean that developers should evaluate the current implementation against their actual requirements.

12. Who should explore OpenDots?

OpenDots may be particularly useful for:

• Developers building custom AI assistants
• Startups exploring AI-powered business workflows
• Teams that want AI tools inside a shared workspace
• Businesses experimenting with computer-using agents
• Developers who want to customize an open-source AI application rather than build every component from scratch

It is especially interesting if you want control over your assistant's roles, tools, interface, and deployment.

Final thoughts

OpenDots represents an important direction in AI applications: moving from assistants that simply answer questions toward assistants that can use tools, interact with software, and help complete real workflows.

Its combination of specialist Dots, document Spaces, computer environments, connected applications, and human approval makes it a useful foundation for developers who want to experiment with persistent AI coworkers.

It is not a plug-and-play autonomous workforce, and some integrations require substantial setup. But if you want to build a customized AI workspace that goes beyond a basic chatbot, OpenDots is worth exploring.

Explore the project: https://github.com/CopilotKit/OpenDots

Source: OpenDots repository and project documentation from CopilotKit.