Introducing CrewContext

In 2024, I started a project called GraphFusionAI.

The idea was to explore how AI systems could coordinate complex processes, especially when multiple intelligent components or agents needed to work together. It started as an experiment around workflow intelligence, system memory, and decision tracking.

Shortly after starting GraphFusionAI, I made a deliberate decision to go quiet.

For most of last year, my focus shifted almost entirely to MsingiAI, where we’ve been working on building AI systems and models designed for African languages and real-world applications. That work has been progressing fairly well and required most of my attention.

But while building and deploying AI systems, one problem kept appearing repeatedly.

When multiple AI agents collaborate on a process, context gets lost between steps. Decisions happen across several agents, but later it becomes difficult to reconstruct what happened, what the system knew at the time, and why a specific action was taken.

That realization led to a shift in direction.

GraphFusionAI has now evolved into CrewContext.

CrewContext focuses on solving a specific problem: preserving context and traceability when multiple AI agents collaborate in a workflow. Instead of agents operating in isolated steps, CrewContext provides a shared layer that records events, decisions, and state changes so the entire process can be reconstructed later.

This matters especially for systems where decisions need to be auditable and explainable, such as financial operations, compliance pipelines, insurance claims processing, and other regulated workflows.

While MsingiAI continues focusing on building AI models and capabilities, CrewContext is emerging as a separate effort focused on infrastructure for multi-agent systems.

We’re still early, but I’m excited to begin sharing more about what we’ve been building and where this direction is heading.

More soon.