Easy-Peasy.AI launches Marky Agent for end-to-end task execution

Easy-Peasy.AI on September 3 launched Marky Agent, an AI system that runs tasks inside its own virtual computer and delivers finished work such as websites, decks, reports and media. The Singapore-based company is positioning the product as a move beyond chatbots toward autonomous workflow execution for individuals and teams.

Why it matters: - Marky Agent is designed to turn prompts into completed work, not just answers. - The product targets workflows that need planning, execution, verification and delivery across multiple steps. - Easy-Peasy.AI is aiming at use cases that matter to teams, including research digests, live websites and recurring reports.

What happened: - Easy-Peasy.AI launched Marky Agent on Sept. 3, 2026, in Singapore. - Marky Agent uses models from OpenAI, Anthropic, Google, DeepSeek, xAI and Meta. - The agent plans multi-step tasks, runs them in its own virtual computer and returns finished outputs. - The launch announcement said Marky can produce deployed websites, presentations, research reports and generated media.

The details: - Each Marky Agent session runs in a dedicated secure cloud sandbox. - The sandbox gives Marky a shell, Python execution, file read and write access, package installation, and chart and document rendering. - Users can watch activity in the “Marky's Computer” panel, which shows browser screenshots, terminal output, search results and files as they are created. - Dmytro Olefyrenko, CTO and cofounder of Easy-Peasy.AI, said each session gets its own sandbox and described the product as more than “a chat window with plugins.” - The company says Marky includes tools and connectors for Gmail, Google Calendar, Google Drive, Notion, Canva and GitHub. - Marky also supports custom APIs and MCP servers. - The launch set includes a full website builder that scaffolds web apps, provisions a database, iterates with live previews and checkpoints, and deploys to a public URL. - Marky can browse the web autonomously, fill out forms and verify its work. - A “take over” mode hands control to the user for logins and other sensitive steps while keeping credentials away from the model. - The agent can generate slide decks, documents, images, videos, music and human-like voiceovers. - Marky supports reusable “skills” that package organization-specific workflows for repeated use. - The system can schedule future or recurring tasks and send results by chat and email. - Marky can create a visible task plan and launch subagents to research and execute in parallel. - The product also includes team collaboration features such as shared chats, shared task plans and team-shared bots and templates. - Marianna Olefyrenko, CEO and cofounder, said customers wanted custom workflows such as weekly market research, landing pages and document-to-deck conversion.

Between the lines: - Easy-Peasy.AI is trying to separate Marky from standard chatbot experiences by centering the product on execution. - The sandbox model also addresses a key trust issue in agentic AI by keeping work isolated and letting users take over sensitive steps. - The broad integration list suggests the company wants Marky to fit into existing work stacks rather than replace them.

What's next: - Marky’s value will depend on how reliably it completes complex tasks across research, creation and deployment. - The scheduling, team collaboration and connector features point to broader adoption in recurring business workflows. - Easy-Peasy.AI says the product is built so users can describe the outcome and have Marky carry the task end to end.

The bottom line: - Marky Agent is Easy-Peasy.AI’s bet that AI’s next step is not better conversation, but finish-the-job automation.