Notes from the team
Writing on agent state, memory for AI tools, and what we're learning while building CTXone.
No posts yet. Subscribe to know when the first one lands.
Watch on GitHub-
From memory layer to context engine
CTXone started as remember / recall / blame. It grew into a four-surface system for everything an agent needs to keep its head straight. Here's the arc.
-
Who told the model that?
AI agents write decisions into your project that no one remembers making. Provenance — ctx blame and why-did-we — is how you get accountability back.
-
The cure for plan rot
Markdown to-do lists start lying the moment work begins. CTXone plans are a first-class object that stays anchored to reality — with proof.
-
CTXone + AgentStateDeveloper: memory meets code intelligence
CTXone remembers what your agents decided. AgentStateDeveloper knows what your code actually does. Together they give an agent both halves of the context it needs.