Ceres: AI Coding Assistant with Voice

Focused AI assistance for the code in front of you.

Ceres Assistant is an AI sidebar for VS Code that helps with code, files, errors, terminal output, Git changes, and debugging work.

It is built for developers who want AI nearby, not AI taking over. Instead of turning every request into a full autonomous run, Ceres keeps the workflow explicit: ask questions, get context-aware suggestions—it is that simple.

Why Ceres?

Many AI coding tools are designed around large agent runs. That can be powerful, but it can also be expensive, noisy, and hard to control.

Ceres takes a smaller, more predictable path:

• Automatically find relevant workspace context, or provide files, selections, errors, terminal output.
• Use OpenAI-compatible APIs, perfect for budget models and local model cases.
• Keep the assistant inside your normal editor workflow.
• Gather only the context that is relevant to your request.
• Get instant suggestions without waiting for a long task to complete.

What Ceres Is

Ceres is a practical AI assistant inside VS Code.

Use it when you want help with:

• Explaining selected code
• Refactoring a function or file
• Understanding an error
• Reviewing Git changes
• Inspecting terminal output
• Troubleshooting debugging issues
• Drafting commands
• Working with local or custom AI providers
• Getting controlled suggestions for the active editor

What Ceres Is Not

Ceres is not designed to take over your whole repository.

It does not try to silently crawl every file, plan a large autonomous task, and rewrite your project on its own. If you want a full end-to-end coding agent that works independently for a long session, tools like Cursor, Copilot agents, Cline/Kilo, Claude Code, Codex, or other agent platforms may be a better fit.

Ceres is designed to understand the relevant parts of your project and provide focused assistance without taking over the entire workflow.

Features

• AI sidebar inside VS Code: Chat with AI without leaving the editor.

• Add selected code or files to chat: Send only the code selection or file you care about.

• Active file context: Automatically read the current file and selection for context.

• Automatic workspace context: Let Ceres find and read the relevant parts of your project when needed.

• Workspace Problems: Send diagnostics and errors from the VS Code Problems panel.

• Git worktree review: Read current staged and unstaged changes.

• Incoming Git overview: Inspect incoming commits and diffs before pulling.

• Git history context: Review recent commits and repository history.

• Voice input: Dictate prompts inside the Ceres sidebar.

• Custom providers and local models: Use OpenAI-compatible endpoints, OpenRouter, DeepSeek, Ollama, LM Studio, and other compatible providers.

• Apply & review suggestions: Apply AI suggestions in all files and review differences before saving. [Pro]

• Run suggestions: Execute AI-generated commands in a context-aware environment. [Pro]

• Editor buttons: Use your own inline buttons and apply AI changes directly to the active editor. [Pro]

• Voice command: Speak about source code and let AI change your code as you describe it. [Pro]

Local And Hosted Models

Ceres is provider-flexible by design. You can use hosted models when you need stronger reasoning, or local and budget models for everyday development work.

Supported providers include:

• OpenAI
• Anthropic
• Mistral
• Mercury
• OpenRouter
• Kilo Code
• Hugging Face
• DeepSeek
• Qwen
• Ollama
• LM Studio
• Minimax (Under development)
• Nvidia (Under development)

This makes Ceres useful for developers who care about cost, privacy boundaries, experimentation, or avoiding lock-in.

Context Control

Ceres gives you control over how project context is gathered and processed.

• A code selection
• A full file
• Current editor context
• Workspace diagnostics
• Git diffs
• Terminal output
• Debugging-related code and errors
• URLs, PDFs, images, audio, and other supported inputs

Better context usually means better answers, lower token usage, and fewer hallucinations.

Navigation Mode

Navigation mode allows Ceres to search the codebase, find and read relevant workspace files, inspect Git information, and gather useful context when needed.

Navigation Mode only gathers information. It does not edit files, run commands, or make changes to your project on its own.

Turn it on or off at any time to continue with the focused workflow, where Ceres uses only the context you provide manually.

Use the Browser and Code Editor Together

Keep Ceres open in your code editor and Chrome at the same time. Your active conversation and current draft update automatically, so you can research in the browser and continue working beside your code without copying and pasting between windows.

Live synchronization requires the browser extension to have access to the same local project folder.

Common Workflows

Explain or Improve a Selection

Select code, run Add Selection to Ceres, and ask for an explanation, refactor, review, or bug check.

Default shortcut:

• Ctrl+Alt+L

On macOS:

• Cmd+Alt+L

Review Before Pulling

Ask Ceres to inspect incoming Git changes before merging them into your working tree.

Edit Code with Voice

Select a code block and describe the change. Ceres rewrites the selected code as you speak.

You can also use speech recognition instead of typing every prompt. Speech recognition helps you describe code changes, debugging notes, and follow-up questions while staying in the flow.

Commands

Ceres contributes these VS Code commands:

• Add File to Ceres
• Add Selection to Ceres
• Add Problem to Ceres

You can access them from editor context menus, explorer context menus, editor title actions, code actions, and the command palette.

Best For

Ceres is a good fit for:

• Developers who want control over AI context
• Teams working on legacy or sensitive codebases
• People using local or lower-cost models
• Developers who prefer assistant workflows over autopilot workflows
• Anyone tired of spending tokens on context that was never needed
• Anyone who wants to edit code with Voice!

Known Limits

• Ceres is not a full autonomous coding agent.
• It can automatically understand your project structure, but it cannot edit files, run commands, or make changes to your project on its own.
• AI suggestions should be reviewed before you use them.

Usage

Ceres is designed for minimal resource consumption and works even with the most affordable models. You are responsible for the provider you connect to, the model you select, and the data you include in prompts. If you use a hosted provider, your data is sent according to their terms. If you use a local setup with Ollama or LM Studio, your code and data never leave your machine.

Bug Reports and Feedback

https://ceres-assistant.com/web/contact.php

Troubleshoot Page

https://ceres-assistant.com/web/troubleshoot.php

Privacy Policy Page

https://ceres-assistant.com/web/privacy-policy.php