✦TL;DR

Assemble, test, and ship the business context every AI agent needs to produce accurate, trustworthy answers.

||

Your AI can read this via Docs MCPcurl -fsSL "https://docs.atlan.com/install-docs-mcp" | bashConnect

Context Engineering Studio

Context Engineering Studio is where you build context repositories. A context repository is the canonical definition of what your data means, scoped to one use case: a portable semantic layer (the metrics, dimensions, relationships, and rules that define your data) that references the Atlan assets it's built from and publishes to each engine in the format that engine expects.

• Deterministic and governed. Every linked asset points back to its governed source rather than being copied, so the same context produces the same repository and stays governed at its source.
• Kept current from the source. When a linked asset changes upstream, the repository flags it so you can regenerate from the current state of that asset.
• One repository, many engines. Snowflake Cortex Analyst, Databricks Genie, dbt, and Claude each read the same model, generated on demand with a review step before you ship.

Read the concept

SALES_PIPELINE

Context repository

Active

Semantic modelStructuredRaw

sales_pipeline

Deal-level pipeline snapshot, one row per opportunity per day.

Dimensions10

Time dimensions3

Facts3

Derived metrics57 not emitted10

Linked assets 4

SALES_PIPELINETable · SALES_DB · PIPELINEReference

OpportunityTerm · Sales glossaryReference

PipelineTerm · Sales glossaryReference

Sales PipelineTerm · Sales glossaryReference

4 linked assets10 derived metrics4 targets

1

Enable

Enable CES on your tenant, assign access to your team, and connect your query engine.

→

2

Assemble

Describe your use case, approve the assets CES proposes, and refine the semantic model. Regenerate when the sources change.

→

3

Deploy

Review what each engine will receive, then deploy to Snowflake Cortex Analyst and Databricks Genie, render for dbt, and expose over MCP for Claude.