Verdent

What is Verdent?

Verdent is an agentic coding suite that turns plain-language requests into planned, reviewed code changes. You describe a feature or product, and Verdent’s Plan Mode clarifies requirements, flags risks and drafts an executable plan before any code changes. It then splits the work into subtasks and runs several coding agents in parallel, each in its own workspace with a separate change history. Beyond code, it can draft documentation and PRDs, analyse data with charts and run research tasks, and newer plans include hosting for deployed apps.

Verdent launched on 23 September 2025. It was founded by Zhijie Chen, formerly head of algorithms at TikTok parent ByteDance, and Xiaochun Liu, formerly head of tech and product at Baidu. The suite comes as a desktop app, first called Verdent Deck, and a Visual Studio Code extension that share credits, memory and settings. In November 2025 the company reported 76.1% pass@1 on SWE-bench Verified using Claude Sonnet 4.5. Verdent was live with a free plan and paid tiers in October 2026.

Key Features

• Plan Mode: Verdent’s Plan Mode turns a vague idea into an executable plan before code is touched. It asks clarifying questions, identifies risks and breaks the job into subtasks, so you approve the approach first.
• Parallel agents: Verdent runs multiple coding agents at the same time on different parts of a project. A backend change, a front-end fix and a test suite can progress together instead of waiting in one queue.
• Isolated workspaces: Verdent gives each parallel code task its own workspace with a separate change history. Native Git worktree support keeps agents from overwriting each other, and you merge only the changes you accept.
• Code review and change summaries: Verdent reviews its own output and presents structured summaries of every code change. A review by The New Stack highlighted DiffLens, which explains why each change was made alongside the diff.
• Full project context: Verdent indexes the codebase to find relevant files and dependencies before it edits. The company says the whole project stays in context, so each new task starts with the full picture.
• Eco Mode and BYOK: Verdent’s Eco Mode runs lower-cost models to stretch credits on routine work. Bring Your Own Key lets you plug in your own model API keys for cost control and closer oversight of data flows.
• Non-coding tasks: Verdent handles documentation, PRD drafting, data analysis with charts and industry research as separate parallel tasks. Product managers and founders can use one tool from specification through to shipped code.
• App deployment: Verdent’s current plans include hosting for deployed apps with project storage and a database per app. The free plan covers one deployed app, while Max allows 50 apps and 32 GB of database space each.

Pros

• Strong benchmark result: Verdent reported 76.1% pass@1 and 81.2% pass@3 on SWE-bench Verified in November 2025. The company says it used its production agent with Claude Sonnet 4.5 and no leaderboard tuning.
• Real parallel work: Verdent runs several agents at once in isolated workspaces, which suits feature-level projects. Teams can split a large change into streams and review each one separately before merging.
• Plan-first workflow: Verdent agrees a plan with you before editing files, which reduces surprise changes. Reviewers noted that it ticks off completed steps and asks permission before multi-file edits.
• Low entry cost: Verdent offers a free plan, a $5 Lite tier and a 100-credit, seven-day trial for new users. You can test it on a real repository before committing to the $19 Starter plan.
• Choice of interfaces: Verdent works as a desktop app on macOS and Windows, a VS Code extension and a JetBrains integration. Credits are shared across the desktop app and extension, so you can switch without paying twice.

Cons

• Credit costs can climb: Verdent bills heavy use through credits, at roughly $0.059 per top-up credit. Large parallel jobs drain monthly allowances quickly, so budgets need watching on Starter and Pro.
• Young company: Verdent only launched in September 2025 and has a short track record. One developer directory reports a team of about 12 people, which may concern enterprises that need long-term support commitments.
• Permissions need clarity: Verdent’s permission settings drew criticism in The New Stack’s review. The reviewer found allow rules and sandboxing of project folders poorly explained, which matters when agents edit files on their own.
• Model details are thin: Verdent says it uses multiple frontier models but names few on its homepage. Buyers who must approve specific AI providers will need to ask the company or rely on BYOK.
• Conflicting pricing listings: Verdent’s plans have changed since launch, and some directories still say it has no free plan or public prices. Check the official pricing page, as third-party summaries may be out of date.

Who is Using Verdent?

• Software engineers: Delegate feature work, refactors and bug fixes to parallel agents.
• Engineering teams: Split large changes into reviewable streams on the Teams plan.
• Founders and product managers: Turn plain-language specs into working, deployed apps.
• Technical writers and analysts: Draft documentation, PRDs and data reports alongside code.

Pricing

Verdent has a free plan, and paid plans cost $5 to $179 a month, with Teams at $20 per user (official pricing page, October 2026).

New users also get 100 free credits for seven days, and top-ups range from 340 to 3,400 credits at about $0.059 each. The bonus credits are listed as a current promotion. Toolradar still says Verdent has no free plan or public prices, which no longer matches the official page. Credit allowances tend to shift with promotions, so recheck the plan details on verdent.ai before you subscribe.

What Makes Verdent Unique?

Verdent treats coding as a team process run by several agents rather than one assistant in an editor. It plans with you first, then dispatches parallel agents into isolated Git worktrees and reviews their output before you merge. The founders come from recommendation systems at ByteDance and Baidu, and the company says it blends multiple models and strategies instead of relying on one foundation model. Built-in app hosting on current plans also takes a project from plain-language idea to a deployed app in one product.

Compatibilities and Integrations

• Platforms: Desktop app for macOS (Apple Silicon and Intel) and Windows.
• IDEs: Visual Studio Code extension and JetBrains IDE integration.
• Version control: Native Git integration with worktree support for parallel changes.
• Models: Multiple frontier models, Eco Mode and Bring Your Own Key options.

Verdent Tutorials

Your first Verdent task follows a plan, run and review loop.

1. Download the desktop app or install the VS Code extension from verdent.ai.
2. Sign up to claim the 100 trial credits, then open your project folder.
3. Describe the feature in plain words and refine the plan in Plan Mode.
4. Let Verdent run the subtasks in parallel workspaces and follow progress.
5. Review the change summaries and diffs, then merge the work you accept.

How We Rated It

Summary

Verdent suits developers and small teams who want to hand whole features to AI agents and review the results, rather than accept line-by-line suggestions. Its plan-first, parallel approach and strong benchmark claim make it a credible rival to single-agent tools. The main limitations are credit costs on heavy jobs and a young company with thin documentation on permissions. Use the 100 trial credits on one real, well-tested repository and compare the merged result with your usual workflow.