We use multiple models to get better results: Claude Sonnet for most of the coding, Gemini Flash for quick tasks and to find relevant files, and we also use a speculative decoding model from Relace AI for fast file rewrites. We use o3-mini as a file editing fallback, and gpt-4o-mini for some trivial tasks.

We also have two modes, which slightly change the above set-up, mainly to achieve higher performance (in max) or lower cost (in lite):

• --max: uses Claude Sonnet and also pulls more files on your codebase to better handle complex problems.
• --lite: uses Claude Haiku for editing instead of Sonnet (~1/3 the cost). It also pulls fewer files, meaning that the context cost will be much smaller.

Codebuff starts by running through the source files in all subdirectories and parsing them with tree-sitter. We use this information to help find relevant files to you requests.

We have a stateless server that passes messages along to Anthropic, OpenAI, or Gemini and websockets to ferry data back and forth to clients. It effectively acts as a proxy between the client and our LLM providers.

We use a fast model to pick the relevant files, and we load them into context Claude Sonnet responds with the right edit. This main agent can also delegate to other agents, as when it does its planning step, or when searching for files.

In general, Codebuff works quite well in larger projects. For example, we've pulled Codebuff into the VSCode repo (millions of lines of code) and have had no issues in our testing. That said, if you're noticing Codebuff getting stuck/forgetting context too soon or using a ton of credits, here are some tips to help you mitigate those issues.

Directory-Based Approach

When working with large codebases, you can start Codebuff from specific directories to maintain focus:

This helps Codebuff concentrate on relevant files and provide more targeted assistance.

Knowledge Organization

For large projects:

• Create separate knowledge.md files in key directories, like so:

  services/
  auth-service/
      knowledge.md
      src/
  user-service/
      knowledge.md
      src/
  api-gateway/
      knowledge.md
      src/
  

• Focus each knowledge file on its directory's specific concerns

• Link related concepts across knowledge files

• Keep root knowledge.md for project-wide concepts

Effective Communication

When working with large codebases:

• Be specific about file locations
• Reference related files explicitly
• Break large changes into smaller, focused requests
• Use the --max mode for better context understanding (--lite tends to have a smaller context window)

Performance Tips

• Leverage your .codebuffignore and .gitignore files to exclude irrelevant directories
• Start in subdirectories when possible
• Break large refactoring tasks into smaller chunks
• Let Codebuff handle file discovery instead of listing all files

Install failed

If npm install -g codebuff gave you an error, try using sudo. Otherwise, try reinstalling node with . Then run the codebuff install command again.

Accessing Your Chat History

Your conversation history with Codebuff is stored locally (and nowhere else) at ~/.config/manicode/projects/<your-project-name>/chats.

If you need help, we'd love if you include a copy of your recent chats for us to help you troubleshoot!

Common Issues

Connection Issues

If you see a message like this:

it means you're having connection issues. In that case, try:

1. Check your internet connection.
2. Try logging out and back in with logout followed by login.

Command Not Found

If you see "command not found" when trying to run Codebuff:

1. Ensure Codebuff is installed globally: npm install -g codebuff
2. Check that your PATH includes npm's global bin directory
3. Try running which codebuff to verify the installation location

Need More Help?

If you're still experiencing issues:

1. Join our for real-time help from other users!
2. Contact us at – please note we prioritize subscribers, but we'll likely reply within a week.

Anthropic recently released Claude Code, their own AI coding assistant.

Both Codebuff and Claude Code:

• Run in your terminal for a seamless coding experience
• Understand your entire codebase context
• Can edit files and execute terminal commands
• Use Claude Sonnet as their primary model

When to Choose Codebuff

Codebuff might be a better choice if you value:

• Speed: Codebuff is
• Cost: Codebuff is one third the cost of Claude Code for equivalent tasks and even less for back-and-forth conversation
• Codebase Analysis: Codebuff pulls more context from scanning your entire codebase, rather than file-by-file. Codebuff also based on their strengths to provide more accurate results.
• Staying in Flow: Codebuff requires fewer confirmation prompts for file edits and command execution.
• Focused changes: Codebuff does just what you asked for, while Claude Code will often get carried away editing more and more files.

When to Choose Claude Code

Claude Code might be a better choice if you:

• Can't use an intermediary provider, need to use the API directly from Anthropic
• Prefer usage-based pricing for occasional use (although this is coming soon to Codebuff!)

Feature Comparison

Summary

Both Codebuff and Claude Code are powerful AI coding assistants that can significantly boost your productivity. Codebuff offers more flexible pricing and performance options, while Claude Code provides direct integration with Anthropic's ecosystem.