Open-Source AI Companion Ankita Releases v2.6 With Windows Computer Use

Overview

Open-source AI developer Ankita has officially released version 2.6.0, introducing native Windows Computer Use capabilities and a deeper integration with existing ChatGPT conversations. The update positions the desktop companion as a more autonomous agent capable of controlling the host operating system directly.

Developed by Sapiens AI, Ankita is designed to bring teammate-style conversations, real tools, and scheduled tasks into a unified desktop environment. The new release also brings refinements to browser automation and fixes critical startup issues identified in the previous 2.5.x series.

The latest update emphasizes autonomous computer control, allowing the AI to interact with Windows interfaces via a visible animated pointer while maintaining the conversational context users expect from a desktop companion.

What’s New in Ankita v2.6

The headline feature of version 2.6.0 is native Windows Computer Use. This capability allows the AI agent to interact with the Windows graphical interface, moving beyond terminal commands and file manipulation to actual screen-based operations.

Key additions in this release include:

• Native Windows Computer Use: The agent can now execute tasks by visually interacting with the desktop environment, complete with a visible animated pointer to provide transparency on where actions are occurring.
• ChatGPT Conversation Integration: Users can now utilize already-open ChatGPT conversations within Ankita. Messages, attachments, and replies remain tied to the selected chat, ensuring continuity with existing model accounts and limits.
• Enhanced Skill Authoring Tools: New tools have been added to facilitate the creation and customization of skills, allowing users to extend the agent’s capabilities more easily.
• Schedule Templates: The release introduces useful templates for scheduling recurring tasks, streamlining the setup of automated workflows.

The architecture separates Core, Desktop, and Plugins connections, enabling ChatGPT to leverage Ankita’s files, commands, browser, and computer controls without compromising the integrity of the original chat session. A notable workflow improvement is the Stop and Resume functionality; stopping ends owned work, while resuming reads the original reply after a restart without redundantly resending it.

Desktop and Companion Refinements

Version 2.6 builds upon significant structural changes introduced in v2.5.0, which completely redesigned the desktop interface. The current iteration offers a compact navigation rail, mascot teammate faces, and responsive work panels that adhere to a consistent graphite style alongside Mono and Slate appearances.

The floating island companion remains a central feature, displaying up to four teammate avatars in two columns. Tool activity rolls upward in real-time, providing live feedback on execution status while full output remains accessible within the chat thread.

Recent patches have also addressed stability concerns:

1. Desktop Startup Repair (v2.5.2): Fixed a critical bug where the desktop failed to start in packaged Windows builds due to a shared browser metadata import reaching host-only code. The security policy remains intact while the import boundary is corrected.
2. Browser Improvements (v2.5.1): Enhanced reliability for form batches and live target checks. The update ensures that mixed fills, clicks, and selects do not invalidate their own references, leading to clearer failure recovery and provider error distinction.

These iterative improvements ensure that the agent maintains accurate state awareness during complex browser automation tasks, reducing confused retries and improving the overall user experience.

How It Works

Ankita operates across three distinct surfaces: a Desktop workspace, a Floating island, and a Terminal CLI. Despite the different interfaces, all three share the same agent runtime and toolset.

The desktop workspace serves as the primary hub for conversations, file review, project management, and connected app settings. The floating island provides quick access to approvals and live tool activity without requiring the main window to be open. Meanwhile, the terminal CLI supports interactive coding, one-shot prompts, and background automation pipelines.

Users can assign a specific working folder and shared context to a teammate by linking it to a project. This allows for specialized personas that understand the nuances of particular repositories or tasks. The system supports a variety of AI providers, including ChatGPT Browser, GitHub Copilot, Groq, Kilo, and any OpenAI-compatible backend.

For those interested in the technical details of the new computer use features or the browser automation enhancements, the complete changelog and verification evidence are available in the project’s official repository. The project remains under the MIT license, encouraging continued community contribution and experimentation.