Puzld.Ai

PuzldAI is a multi LLM orchestration framework that wraps existing command line AI tools, including Claude, Gemini, Codex, Ollama and Mistral. Because it drives those installed CLIs, the documentation states that no separate API keys are required by PuzldAI itself. It is used from a terminal interface or through commands such as puzldai run, puzldai compare and puzldai agent. The framework targets developers who want to use several models together instead of committing to one. A router analyzes each prompt, decides the task type and sends it to the most suitable agent, using a local Ollama model for routing by default and falling back to a configured agent, Claude by default, if a call fails. Multi agent modes let models compare answers, debate, correct one another or reach consensus. Pipelines and workflows chain agents together, and an Autopilot mode runs tasks autonomously. An agentic mode gives models tools to explore a codebase, plan, edit files, run bash commands and generate code, which the documentation likens to Claude Code but usable with any supported model. Memory and RAG features provide persistent memory with vector search and automatic context injection, and indexing features add AST indexing, dependency graphs and semantic code search. The framework can also generate training data. The documentation site publishes no pricing or plans. Underlying model usage is governed by whichever CLI tools and providers the user already has.