This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
This is an AI-powered coding assistant built with the Agentuity platform. It provides a hybrid cloud-local architecture where AI logic runs in the cloud while file operations and code execution happen locally for security.
# Development
bun run dev # Start local development server (port 3500)
bun run build # Build using Agentuity bundler
bun run start # Bundle and run locally
# Code Quality (ALWAYS run before completing tasks)
bun run format # Format code with Biome
bun run lint # Lint code with Biome
# CLI Testing
bun run cli # Run CLI in development mode
bun run cli --local "test message" # Test local agent
bun run test-config # Test dynamic configuration
bun run show-urls # Show detected agent URLsThe project uses a hybrid cloud-local architecture:
- Cloud Agent (
src/agents/CloudCoder/): Contains the AI logic that runs on Agentuity cloud or locally - CLI Client (
cli.js): User interface that connects to the agent and renders markdown - Tool Proxy (
cli/tool-proxy.ts): Executes tools locally based on cloud agent requests - Continuation Handler (
cli/continuation-handler.ts): Manages tool execution workflow
All tools are defined in tools/interface.ts with shared schemas between cloud and local components.
-
Tool Development: When adding new tools, update both
tools/interface.ts(schemas) andcli/tool-proxy.ts(implementation) -
Security Model: Only whitelisted commands can be executed locally. See
allowedCommandsincli/tool-proxy.ts -
TypeScript Configuration: Strict mode is enabled. All code should follow strict typing rules
-
Code Style: Use Biome for formatting and linting. No manual formatting changes needed
-
Agent Configuration: The
agentuity.yamlfile is auto-generated. Never edit it directly
- No dedicated test framework is configured
- Test manually using
bun run cliwith various commands - Use
bun run test-configto verify configuration detection - Test both cloud and local modes during development