Puzld.Ai

PuzldAI is a multi-LLM orchestration framework that wraps existing CLI tools such as Claude, Gemini, Codex, Ollama, and Mistral — no API keys required. It enables users to route tasks to the best agent, compare responses, chain agents in pipelines, or let them collaborate in multi-agent modes. The framework includes an agentic mode that gives LLMs tools to explore codebases and propose edits, similar to Claude Code but compatible with any LLM. Key capabilities include workflows and automation (pipelines, autopilot), multi-agent modes (compare, debate, correction, consensus), intelligent routing based on task complexity, agentic execution (plan and build modes with file editing and bash commands), persistent memory with vector search and automatic context injection, and indexing/search features (AST indexing, dependency graphs, semantic code search). PuzldAI is designed for developers and teams who want to leverage multiple LLMs without managing separate API keys or accounts. It supports both TUI commands (e.g., /compare, /debate, /autopilot) and CLI commands (e.g., puzldai run, puzldai agent). What sets PuzldAI apart is its ability to orchestrate multiple LLMs alongside existing CLI tools, providing a unified interface for complex workflows. It is particularly suited for tasks like code generation, debugging, research, and training data generation, all while maintaining flexibility and control.