We built it because we needed it.
ContextWorks started inside RevWorks — the thing we needed before we could build and scale AI features users trusted.
AI context needs architecture.
Every AI feature we built landed on the same requirement underneath — and every agent that came after would land there too. Context wasn't going to be a feature; it was the layer everything else would be grounded on. So we stopped building it per feature and built the layer they could all share.
Not a model. Not a prompt.
What every feature needed first was a trustworthy view of the data underneath — which fields matter on this object, what they mean, which related records travel with them, and absolute certainty that the person asking sees only what they're allowed to see.
Three features in, the pattern was clear.
Chat, the daily advisor, and deal assessments each needed the same trustworthy view, assembled the same way. Most of the time this ends up as a field list into each prompt — which is how you end with four definitions of the same truth.
So we built it once, as configuration. One engine, one definition, every surface.
From the foundations of RevWorks.
Akoonu has been a Salesforce ISV partner since 2017, building native apps for revenue teams. ContextWorks came out of the AI foundation we built inside RevWorks — designed to extend across features, and across surfaces.
Salesforce ISV partner
Akoonu starts building native applications for revenue teams — forecasting, pipeline reviews, sales insight. Everything inside the customer's org.
AI foundations in RevWorks
Chat over pipeline data, a daily advisor, deal assessments. Each one needed a trustworthy view of CRM data before it could say anything useful — and the first configuration interface for choosing those fields was built.
MCP and Agentforce surfaces
RevWorks features and data services become reachable from outside the app. The same question starts arriving from more directions, and the context has to match — or the answers won't.
The entity assembly engine
All feature-based context code collapses into one declarative engine: entities defined once in a builder, assembled live at call time. Abstracted deliberately — built to answer whatever an agent asks on whatever surface, not the feature it came from. It ships inside RevWorks.
ContextWorks, standalone
The same engine, lifted into its own managed package — installable in any org, on any surface, for any agent.
