Cellium Agent is a lightweight Python AI agent framework for building autonomous agents with adaptive decision-making. It uses a local-first, minimal architecture instead of heavy external dependencies or complex abstractions, making it easy to understand and extend. Key features: - Adaptive decision loop (Plan-Execute-Observe-Replan) - Bayesian bandit-based strategy optimization - Three-layer memory (persona, session, long-term) - Event-driven microkernel design - No required vector database
No reviews yetBe the first to leave a review for Cellium Agent
Maker
📌
I built Cellium Agent after being frustrated with how heavy and fragmented most AI agent frameworks are.
Many tools today either depend on external services or introduce too many abstractions, making it hard to understand what is actually happening in the agent loop.
The goal with Cellium was to keep everything local, minimal, and explicit — where the full decision cycle (plan, execute, observe, replan) is visible and controllable.
Would love to hear feedback from others building agents: what do you think is still missing in current agent frameworks?