Overview

LevelCode agents spawn other agents, share tools, and pass context between tasks:

• Code Generation - writes code
• Review - catches bugs and style issues
• Research - finds docs and examples
• Planning - breaks down requirements
• File Discovery - navigates codebases

Control agents with TypeScript generator functions. Orchestrate workflows, branch on file contents, and make steps deterministic instead of hoping a prompt lands the right way.

• levelcode/base - Main coding assistant (orchestrates other agents)
• levelcode/editor - Code editing and file modifications
• levelcode/reviewer - Code review and feedback
• levelcode/thinker - Deep thinking and problem analysis
• levelcode/researcher - Web search and documentation lookup
• levelcode/file-picker - File discovery in your codebase
• levelcode/commander - Terminal command execution
• levelcode/code-searcher - Search patterns in code files

Browse all available agents at the Agent Store.

A typical call to LevelCode may result in the following flow:

If you ask "refactor this authentication system", LevelCode might:

1. File Picker finds auth-related files
2. Research looks up best practices
3. Planning creates step-by-step plan
4. Base implements changes informed by the previous agents
5. Reviewer checks for security issues

Agents coordinate through the spawnerPrompt field, which tells others when to spawn them. Agents can spawn others listed in spawnableAgents.

Start LevelCode and run the /init command to set up your project:

levelcode

Then inside the CLI:

/init

1. Customize existing agents - Modify prompts and tools
2. Create new agents - Build specialized functionality
3. Reference guide - Complete field documentation