Bridgic

Bridgic is an open source Python framework for building AI agents and LLM applications. Its central idea, summed up in the tagline "Bridging Logic and Magic", is that deterministic workflows and autonomous agents should run on one runtime rather than two separate stacks. Both are orchestrated through a Dynamic Directed Graph engine, so a developer can start with explicit, predictable control flow and hand selected steps to a model that decides which tools to call and which path to take. The framework is aimed at Python developers who need agent behaviour inside systems that must stay reliable. The graph can change its own topology and routing at run time, components can be nested and reused as modules, and parameter resolving passes data between workers without global state. Developers choose between a declarative API and Agent Structure Language, a Python native DSL for describing agentic structures in little code. Human in the loop support lets a run pause for asynchronous feedback, or be interrupted and resumed later with its state persisted. Model integration works with any LLM provider through one abstraction, MCP servers can be connected so their tools act as first class workers, and tracing plugs into observability platforms such as Opik and LangWatch. Bridgic is free. It is released under the MIT licence, installs from PyPI with pip or uv, and requires Python 3.9 or newer. There is no paid tier or hosted service on the documentation site. The project is developed under the bitsky tech organisation on GitHub, and its documentation includes tutorials plus an API reference covering the core automa, model, prompt and configuration packages.