Verdent AI Inc., an artificial intelligence startup developing agentic coding tools, today announced the launch of its platform that can plan, self-verify and iterate on complex software development work by guiding multiple agents in parallel.

The company was founded by Zhijie Chen, former head of algorithms at ByteDance Ltd.’s short-form video platform TikTok and senior executive from Chinese multinational company Baidu Inc. Verdent focuses on building its platform to tackle large-scale, long-horizon tasks that current agentic AI coding suites may struggle with.

Agentic coding represents a trend in software development tools that employs autonomous AI agents to extend beyond AI-assisted autocomplete. AI agents can interpret user goals written in plain language, plan out new features or entire projects, make decisions on how to approach them and break them down into sub-tasks. That allows them to write, modify and maintain code with little or no human intervention.

Verdent is available in two formats: a desktop coding agent orchestration platform called Verdent Deck and a plugin for Microsoft Corp.’s VS Code development environment. The standalone desktop interface is now available for macOS, with a Windows release planned soon.

“Verdent and Verdent Deck mark the move from keystroke-level completions to true outcome-driven delegation,” said Chen, who’s also chief executive. “Our goal is to be the human developer’s enabler in the AI era. Agents handle the repetitive work. People direct strategy, validate results and build what’s next.”

What sets Verdent apart from the variety of agentic coding platforms currently on the market is its ability to work with developers through planning and verification. Developers can set the platform to “planning” mode, which can take vague descriptions and turn them into a requirements document and action plan. From there, developers can modify the action plan to fit their intent before unleashing multiple AI agents to complete the work.

Verdent Deck launches multiple agents to work in parallel, each isolated in its own Git-enabled codespace and virtual environment. The platform orchestrates the agents and keeps them separate with task isolation, allowing them to complete work faster without colliding with one another while writing code.

Once the process is complete, agents will provide a change summary and a version difference report. That gives users transparency into the AI’s work. Users can easily use this documentation to create patch notes for their audience or review it to determine their next steps.

In the Verdent Deck interface, developers can commit code, open pull requests or roll back changes that they didn’t want made. According to Codeck, it’s designed to allow software developers to focus on task assignment, result validation and producing quality code with minimal intervention.

The Verdent plugin for VS Code offers an interactive interface for advanced collaboration with agents within a coding environment. After completing and validating work, developers can dive directly into software changes and make their own modifications between prompting AI agents to produce code.

Verdent is entering a market saturated with agentic AI coding tools, including Anysphere Inc.’s Cursor and Cognition AI Inc.’s Devin. Technology mega giants Google LLC and Amazon.com Inc. have also released autonomous coding agents, Jules and Amazon Q Developer, respectively.

Verdent is available for developers via a subscription with a free trial starting at $19 per month. Developers are provided a certain number of AI credits per month, limiting the number of agent calls that can be made. Every plan allows users to pay extra for more credits. The basic package includes 340 credits, and an additional 240 credits can be purchased for a starting price of $20.

Image: SiliconANGLE/Microsoft Designer