VIDA — AI Coding Agent for VS Code

Build, understand, and improve software without leaving Visual Studio Code. VIDA can search your workspace, explain code, review changes, edit files, and run approved development commands—with clear controls over what it can do.

Cloud models: GLM 5.3, GLM 5.3 Flash, GLM 5.2; Kimi K3 and Kimi K2.7 Code; DeepSeek V4 Pro, DeepSeek V4 Flash, DeepSeek V4.1 Flash, DeepSeek V3.2, DeepSeek Coder, DeepSeek V3, and DeepSeek Reasoner; and MiniMax M3. Cloud models are offered across all four plans—Free (limited), Starter, Pro, and Ultra. Model access and usage allowances vary by plan.

Use your own local model: connect Ollama, LM Studio, or another OpenAI-compatible endpoint. Local requests go directly to the endpoint you configure and do not use VIDA cloud credits.

To open VIDA: Open the Command Palette (Cmd/Ctrl+Shift+P), choose View: Open View..., and select VIDA. You can also open it from the VIDA icon in the Activity Bar. If the icon is hidden, enable Activity Bar under View → Appearance or find VIDA in the hidden views menu.

Get VIDA · Plans and cloud access · Support

Pricing at a glance: local-model requests do not use VIDA cloud credits · cloud models with a Vidlogi plan — see plans.

Why VIDA

• VIDA Team mode: a fast, low-cost model does the work while stronger advisor models review plans and the final changes — high quality without paying for a large model on every step.
• Undo file edits: VIDA checkpoints edits made with its file editing tools. Review the diff before approval and use the checkpoint to restore those files.
• Built-in skills and slash commands: 17 expert playbooks (code review, debugging, tests, refactoring, security, performance, migrations, docs, and more) plus quick commands such as /review, /fix, /test, /explain, /plan, /commit and /init. Add your own skills per project.
• Project-aware by default: a local project index, language-server navigation, @file and @problems mentions, file attachments, and project rules/memory files so answers fit your codebase.
• Analyze screenshots and images: attach an image and use a vision-capable model, such as GLM 5.3 Flash, to ask about its contents.
• Generate images from chat: ask VIDA to create an image; it requests approval for the displayed media-credit cost, then saves the result in your workspace.
• Produce and edit PDFs: subscribers can create book interiors from chapter files, add title/copyright/contents pages, set trim size and mirrored margins, and add page numbers and running headers. Create a full-wrap cover using printer-provided dimensions and optional artwork, or use the other PDF tools to crop and fit images, merge or split PDFs, reorder/rotate/delete pages, fill standard forms, add captions and watermarks, and set metadata. PDF processing runs locally; generated files are written after approval.
• Cloud or local, your choice: use VIDA cloud models from leading open model families such as DeepSeek, GLM and Kimi (availability depends on your plan), or connect Ollama, LM Studio and other OpenAI-compatible local models. Local-model requests do not use VIDA cloud credits.
• Works in your language: the interface is available in 15 languages and VIDA replies in the language you write in.

What VIDA can do

• Understand a codebase: search project files, inspect relevant code, and answer questions with workspace context.
• Create visual assets: generate an image from a prompt and save it into the current project after you approve the credit charge.
• Create and organize PDFs: the PDF workbench is available to active paid subscribers. It can lay out chapter files as a book interior, create a full-wrap cover from supplied printer dimensions, make a photo book from ordered workspace images/text, and perform common PDF and page operations locally. PDF processing does not use VIDA cloud credits; drafting text with an AI model follows that model's normal account and usage rules.
• Help implement changes: create and edit files, explain selections, and review the resulting diff.
• Work with developer tools: run terminal commands with approval, inspect Git changes, and use MCP servers you configure.
• Browse from chat: use VIDA's browser, or enable Computer Control on macOS or Windows to use a browser window you already have open. Reading and interaction depend on what the browser exposes through accessibility.
• Fit your workflow: keep searchable conversation history, save scheduled tasks, and use VIDA in English, Arabic, Simplified Chinese, Traditional Chinese, French, German, Spanish, Italian, Brazilian Portuguese, Russian, Turkish, Japanese, Korean, Polish, or Czech.
• Choose cloud or local AI: cloud use follows your Vidlogi plan. Local-model requests do not use VIDA cloud credits when a compatible local endpoint is running.

Safety and control

Choose which file, command, and browser actions VIDA can run automatically in your user settings, and stop a task whenever you like. Computer Control is off by default. VIDA requires a trusted workspace and asks for approval before launching an app or capturing your screen.

Get started

1. Install VIDA from the Visual Studio Code Marketplace.
2. For cloud models, choose Sign in with vidlogi.com and complete Connect VS Code in your browser.
3. For local models, start a compatible local model server and select its model in VIDA. Local requests do not pass through the VIDA cloud gateway.
4. Ask a project question such as “Explain the request flow in this workspace” or “Review my uncommitted changes.”

Privacy

Cloud requests are sent through the Vidlogi service, which keeps model-provider credentials server-side. Local-model requests are sent to the endpoint you choose. If you ask VIDA to read an open window, the text it reads can be included in requests to the model you selected. Conversation history is stored on your device and separated by signed-in account on that device. Sign-in uses an authorization flow, and the resulting credential is stored in VS Code SecretStorage.

VIDA asks before enabling optional usage analytics, which are off by default. If you allow it and VS Code telemetry is enabled, VIDA sends limited events to Vidlogi: extension activation, successful sign-in, chat start/outcome, operating system, extension version, and whether a local or cloud model was selected. When signed in, the existing authorization credential lets the service associate events with your account; analytics records store a pseudonymous hashed account ID, not your email. Telemetry events do not include prompts, source code, file paths, or model names. Turn it off in Settings → VIDA → Telemetry Enabled, or disable telemetry in VS Code to stop these events.

Review the privacy information and terms before using cloud services. Model availability and cloud usage depend on your account and plan.

Compatibility

• Visual Studio Code 1.90 or later
• macOS, Windows, or Linux for the VS Code extension
• Chrome for VIDA's controlled-browser feature
• Optional: Ollama, LM Studio, or another compatible local/OpenAI-style model endpoint

FAQ

Does VIDA send my code anywhere? Only what a request needs. With cloud models, the relevant context is sent through the Vidlogi service to the model; with local models, requests go only to the local endpoint you choose. Credential files such as .env are never read or sent.

Can I use a local model offline? Local inference can work offline when a compatible model and endpoint are installed and running. Cloud models and online integrations need an internet connection.

Can I chat with a local model without an account? Yes, when a compatible local model endpoint is running.

How do I stop VIDA? Press Stop in the chat or the Stop VIDA button in the status bar at any time. Every optional capability (browser, terminal, MCP, computer control, and more) can also be turned off in Settings → Features.

Can I undo changes? Yes. Use Undo these changes after a request, or VIDA: Restore Checkpoint… to roll back several requests.

What PDF and book tools are included? The paid PDF workbench supports book interiors from chapter files, optional front/back cover artwork, print layout settings, and common PDF operations. Cover dimensions such as spine width must come from your chosen printer. It does not currently OCR scanned pages, replace arbitrary existing page text, or open password-encrypted/XFA documents. PNG and JPEG images are supported. Provide a local TTF/OTF font for non-Latin text; complex right-to-left shaping should be checked in your PDF viewer.

Support

Visit vidlogi.com or email support@vidlogi.com.

Chat stays flexible: local-model conversations remain available without signing in. The separate PDF workbench requires an active paid Vidlogi subscription.

© 2026 WestXPro Ltd · vidlogi.com