Overview
Demonstrates how agents can autonomously search codebases by:- Searching for files matching patterns
- Reading relevant files
- Synthesizing information into answers
- Using subagents for specialized tasks
SearchFilesTool: Search for files by pattern or contentReadFileTool: Read file contentsUseSubagents: Spawn specialized subagents for subtasks- AgentRegistry: Registry of available subagent specifications
- Autonomous search: Agent determines strategy based on question
Example
Expected Output
Key Points
- Autonomous search: Agent determines which files to search for and read
- Subagent orchestration: Main agent spawns specialized searcher and reader subagents
- AgentRegistry: Registry of available subagent specifications
- Specialized tools: Each subagent has specific tools (search, read)
- Multi-step reasoning: Agent synthesizes information from multiple file reads
- Dynamic strategy: Agent adapts search based on initial findings
- Tool chaining: Search results inform which files to read
- Use cases: Code documentation, architecture analysis, dependency mapping, test coverage reports