Bytebot

Market position

Bytebot presents itself as an open-source container for AI desktop agents and a next-generation replacement for robotic process automation, competing with script-based RPA suites such as UiPath and Automation Anywhere, with Microsoft Power Automate Desktop, with browser-only agents and cloud browser infrastructure providers, and with general AI agent frameworks. Y Combinator categorizes the company under generative AI, robotic process automation, SaaS and AI assistants. The project's open-source repository has drawn roughly 11,100 stars and 1,500 forks.

Bytebot's stated distinction from browser-only agents and API integrations is that the agent controls an entire desktop, letting it use email clients, office tools, IDEs, command-line utilities and any software the user installs, and bridging legacy systems that expose no API. Compared with traditional RPA platforms such as UiPath, the company positions the product as requiring no flowcharts or scripts, adapting to interface changes because it interprets the screen visually rather than relying on fixed selectors, and handling unexpected popups or variations. Self-hosting is presented as an additional differentiator: tasks and data remain on customer infrastructure, each desktop is isolated in its own container, and customers use their own LLM API keys without platform usage limits. The codebase is open source and portable across local Docker, Railway, Kubernetes and major cloud providers.

Technology

The stack combines an Ubuntu 22.04 virtual desktop running XFCE4 with pre-installed applications (Firefox, VS Code, Thunderbird) and a bytebotd automation daemon; a NestJS agent service that calls Anthropic Claude, OpenAI GPT or Google Gemini to plan and execute steps; a Next.js web application for task creation and live desktop viewing; and REST endpoints for task management and low-level computer-use actions such as screenshot capture and click_mouse at given coordinates. Files can be uploaded to the desktop and read into the model context. Deployment options include Docker Compose, Railway, Helm charts for Kubernetes, and AWS, GCP or Azure. The project supports the Model Context Protocol and its documentation is exposed through an llms.txt index.

Go-to-market

Distribution is developer-led and open source: a public GitHub repository with quick-start instructions promising deployment in about two minutes, a Railway one-click template, documentation, a Discord community and an X account. The website publishes recorded demonstrations of the agent handling 2FA logins, development workflows and technical research and summarization, plus a blog of technical posts. The earlier browser-automation product was sold through self-service signup with tiered pricing for individual engineers and companies, supplemented by booked calls with the team.

Developers building computer-use agents, plus engineering, operations and back-office teams at companies automating processes that span multiple applications. Documented workloads span finance and banking operations, compliance and regulatory filing, vendor and invoice management, purchase order and ERP entry, HR record keeping, customer onboarding, QA and UI testing, and research and competitive analysis. Self-hosting is aimed at organizations with data-privacy requirements that preclude sending task data to a third-party cloud.

Geography

The company is listed as based in Boston. Because the software is self-hosted, deployments run wherever customers operate their own infrastructure, from local Docker environments to Railway, Kubernetes clusters and AWS, GCP or Azure.

History

The company was founded in 2021 and took part in Y Combinator's Summer 2021 batch, based in Boston with a team of three. Its initial product, presented in a YC launch post, focused on natural-language browser automation for web scraping and form filling, including a hosted browser API and integrations with existing Puppeteer and Playwright setups. The product later evolved into a full desktop agent running in a containerized Linux environment; company blog posts published in April, June and July 2025 chart this shift, covering the case for desktop agents, the choice of a simple desktop-agent abstraction, and the Bytebot Core container as an agent control surface. Y Combinator lists the company's status as acquired. The main open-source repository, bytebot-ai/bytebot, was archived by its owner on 7 March 2026 and is now read-only.

Risks & controversies

The primary open-source repository was archived and set to read-only on 7 March 2026, ending public development in that repository. Because the agent operates a live desktop with access to password managers, browser sessions and uploaded documents, deployments involve handling credentials and sensitive files; the company addresses this by keeping execution on customer-controlled infrastructure and isolating each desktop in its own container.