CopilotKit has introduced OpenBot 1.0, an AI-powered workspace designed to help businesses automate tasks while maintaining control over how artificial intelligence agents operate. Released on November 9, the platform allows organisations to run AI agents on their own infrastructure and select the models, agents, and tools that best fit their requirements.
The launch highlights a growing shift towards AI systems that can handle practical workplace responsibilities rather than simply answering questions. OpenBot is designed to let users assign tasks, monitor an AI agent’s progress, review its findings, and approve important actions through a single conversation interface.
One of the platform’s key features is its focus on user control and security. Instead of allowing AI agents to perform every action automatically, OpenBot uses a policy gateway to check operations before they are executed. These actions can include browser clicks, file modifications, and calls to external tools.
How OpenBot 1.0 Works
OpenBot provides a workspace where users can interact with individual Bots and assign them different responsibilities. After receiving a task, a Bot can work within a browser environment, access permitted tools, and use designated workspace folders.
The platform uses Chromium browser profiles to support browser-based activities. Businesses can also provide individual Bots with isolated containers, helping separate their working environments and resources.
Users can follow the progress of assigned tasks, examine the results, and decide whether the Bot should proceed with additional steps. This approach is intended to make AI automation more transparent by keeping people involved in decisions that require approval.
OpenBot 1.0 also supports flexible output formats. Markdown is used for responses by default, while the workspace can present information through charts, checklists, tables, and cards when those formats are more suitable.
Another important aspect is the ability to choose which models, agents, and tools are available. This gives organisations greater flexibility when building AI-powered workflows around their existing requirements.
Security and Approval Controls
A major feature of OpenBot is its policy gateway, which evaluates an agent’s proposed actions before allowing them to proceed.
The system uses Common Expression Language (CEL) rules to assess different operations, including browser interactions, file writes, Model Context Protocol (MCP) calls, and interface components.
According to the platform’s described approach, deny rules are evaluated before other permissions. The gateway records its decision before an approved action is carried out. If a required policy is missing or a rule fails, the system blocks the action.
These controls are intended to reduce the risk of AI agents carrying out unauthorised operations, particularly in business environments where access to company files, communication platforms, and other tools must be carefully managed.
Potential Uses for Businesses
OpenBot’s design supports several workplace automation scenarios. For example, a Bot could organise customer support tickets and prepare them for further processing while waiting for human approval before publishing a message to Slack.
It could also identify expenses that exceed company policy limits or prepare suggested software fixes for a developer to review before implementation.
These examples demonstrate how businesses could use AI agents to handle repetitive work without giving them unrestricted authority. Employees can delegate suitable tasks while retaining oversight of sensitive decisions and external actions.
MIT Licence and Self-Hosted AI Workflows
OpenBot 1.0 is released under the MIT licence, giving organisations the flexibility to use and adapt the workspace under the licence’s terms. Its self-hosting approach also allows companies to operate AI agents on their own infrastructure rather than relying entirely on an externally managed environment.
The ability to select models and tools, configure agent permissions, and apply policy-based approval checks makes the platform relevant to businesses exploring controlled AI automation.
As organisations increasingly experiment with AI coworkers, platforms such as OpenBot reflect an approach that combines task automation with human supervision. Its focus is not only on what an AI agent can accomplish, but also on defining which actions it is allowed to perform and when a person must approve them.