All You Can Data
Ace turns your governed semantic layer into agent-ready context. Point any agent at a single audited source, and Ace serves your models, metrics, and join paths as grounded context, never raw data. Many agents, one governed context.
Grounds every agent
Claude
Cursor
OpenAI
Exmergo Viz
Snowflake
Databricks
BigQuery
PostgreSQL
DuckDBExmergo DexGoverned
Semantic Layer
Single source of truth
Exmergo AceAgent Context Engine
Exmergo VizBusiness users · Self-serve dashboards
Your custom agent · Build on the same engine
Sources
SnowflakeDatabricksBigQueryPostgreSQLDuckDBExmergo DexGoverned
Semantic Layer
Single source of truth
Exmergo AceAgent Context Engine
Serves
Exmergo VizBusiness users · Self-serve dashboards
Your custom agent · Build on the same engine
The Agent Context Engine
Agents are only as good as the context they stand on. Point them at raw SQL and they invent joins, miss metric definitions, and surface data they were never meant to see. Ace sits between your governed semantic layer and every agent that consumes it, serving grounded, permission-aware context so answers stay correct, consistent, and auditable, no matter which agent asks.
One interface, every agent
Ace is the single context interface every agent connects to. Exmergo Viz, your own custom agents, and the coding agents your team already uses all read from the same place.
- •One connection point for Viz, custom agents, and coding agents
- •No bespoke integration per agent: connect once, ground everywhere
- •Add a new agent without re-plumbing your data access
Grounded context, not raw data
Ace serves your dbt models, metrics, and join paths as structured, agent-ready context. Agents reason over definitions instead of guessing at schemas (or even worse, from raw data)
- •Metrics and dimensions served with their real definitions
- •Verified join paths, so agents never invent relationships
- •Answers that match the numbers your team already trusts
Governed by default
Opening your data to agents should not mean opening it up. Ace enforces your access model on every request, so each agent sees only what it is allowed to.
- •Role-based access control and workspace-level permissions
- •PII policy enforcement on every agent query
- •Full access audit trails, from prompt to answer
Always in sync
Ace grounds on the same governed layer Dex maintains, so agents never work from a stale copy. When the semantic layer changes, every agent sees it.
- •One source of truth shared by every consumer
- •Definitions stay current as Dex authors and maintains them
- •No drifting exports or duplicated modeling per agent
How Ace works
One governed layer in, agent-ready context out.
Point Ace at your governed layer
Ace connects to the dbt models and semantic layer Dex maintains as your single source of truth.
Ace serves agent-ready context
Definitions, metrics, and join paths become grounded, permission-aware context, never raw SQL.
Every agent queries one source
Viz, your custom agents, and coding agents all ground on the same audited layer, under the same RBAC, PII policy, and audit.
Governed from the first query
Ace applies your access model to every agent request, so grounding more agents never widens your exposure. Security and control are built in from the start.
- Role-based access control (RBAC) and workspace-level permissions
- PII policy enforcement on every agent request
- Full access audit trails for every agent and query
- Clear lineage, from governed model to agent answer
Become an Ace design partner
Ace is in private preview. We are working with a small group of design partners to ground their agents on one governed context layer, and shaping Ace around what they need.
Apply as a design partner