Developer Ilia Kuzmin has released DelphiLoop, an agentic AI coding application built entirely in Delphi that uses two AI models in a continuous generate > review > refine loop to produce higher-quality Object Pascal code.
Instead of relying on a single LLM response, DelphiLoop assigns one model as the Executor to write code while a second Reviewer analyzes it for bugs, memory leaks, logic errors, and poor practices. The cycle repeats until the reviewer reports no remaining issues or a configurable iteration limit is reached.
DelphiLoop works with local models through Ollama as well as cloud providers including OpenAI, OpenRouter, and other OpenAI-compatible APIs. Models can be mixed and matched, allowing, for example, a local coding model to collaborate with a cloud-based reviewer.
The application also includes a native FireMonkey chat interface, Delphi syntax highlighting, collapsible code revisions, customizable agent prompts, token usage tracking, and one-click code copying—all implemented without browser components or external dependencies.