Densery
Densery Platform · Semantic layerThe guide book your agent reads first.
Which systems exist and what each is for. What every table and field means, how it aggregates, how it joins. The words your people use. The tools an agent may call. Tool-call accuracy from ~70% to 98%.
A new hire reads the company handbook before touching a system. So does an agent — if you write one.
One guide book. Systems, data, vocabulary, tools.
What each system is for. What every table means and how it aggregates and joins. The terms your people use. The tools an agent may call — served as a menu of at most ~30 items.
Daily revenue per outlet, after returns. One row per outlet per day.
From ~70% before the guide book existed.
What the semantic layer does.
Systems, explained
Core banking, claims, CRM, ERP — what each is for, what lives where, which one is the source of truth for what.
Every table, described
Business meaning per field, the grain of each table, how it aggregates and how it joins — so an agent queries like a good analyst.
Vocabulary and rules
The words your people use — “active customer”, “outlet above plan” — defined once and reused by every agent.
Tools as a menu
At most ~30 items a level, described like an onboarding guide, tuned per tenant. Hidden is not denied: every call is still authorized, fail-closed.
Where this shows up in the work.
Connect your AI
Point Claude Desktop or any MCP client at Densery and it inherits the map, the tools and the guardrails in three steps.
See the runtime →Measure the menu
Evals compare two curations on the same golden set, so a wording change is a measured change.
See governance →Traces
See what the agent actually called, in what order, and what it cost.
Open governance →Thirty minutes with Densery.
Tell us about the work you want an agent to finish. We listen first, then show you the platform and the case that is closest to yours. This is a working session, not a sales pitch.