Documentation index: llms.txt. This page is also available as markdown: append .md to this URL or send Accept: text/markdown.
How Coalesce Fits Together
Speed isn't the problem, coordination is. Coalesce unifies data transformation, cataloging, and monitoring into one governed platform, so humans and AI agents build data to the same standard, in the same flow, with the same context.
This page explains what that means in practice: what each product owns, how they share metadata, and the ways people and agents work with them.
One Metadata Layer, Three Products
| Product | It owns | It reads from the others |
|---|---|---|
| Transform | How data is built: Nodes, Node Types, column-level lineage, tests, Environments, and deploys. | Descriptions written in Catalog, synced back to your Nodes. |
| Catalog | What data means: discovery across your stack, documentation, ownership, and usage. | Which tables Transform built, their lineage, descriptions, and test results. |
| Quality | Whether data is healthy: monitors, tests, and lineage-aware alerting. | Lineage, so an alert shows what is affected downstream and who owns it. |
Ways to Work
Every path below reads and writes the same metadata, under the same permissions. Work started in one place is visible in the others.
| You work | With | Best for |
|---|---|---|
| With your AI agent, locally | Your pipeline as files in Git, the coa CLI, Coalesce Desktop to review, and Coalesce platform skills to teach your agent | Engineers who build with an agent like Claude Code, Cursor, or Codex. See Build With Your Agent. |
| In the web app | The visual builder, the mapping grid, and Coalesce Copilot | Building and editing in the browser with no install. |
| Through MCP | Transform MCP and Catalog MCP | Agents that check runs, inspect deployed Environments, or look up what a table means. |
| Through APIs and the CLI | The REST API and coa | CI/CD, orchestration, and custom integrations. |
Agents follow the same rules as people. An agent can only do what the account it runs as can do, and its changes go through the same Git history, deploy plan, and review as an engineer's.
One Change, Followed Across the Platform
Here is what happens when someone adds a customer dimension.
- Build. Your agent writes
DIM_CUSTOMERas a SQL Node and runs it in your development schema. You review the graph, columns, and data in Coalesce Desktop, then commit. - Ship.
coa planshows what will change in the shared Environment. After you approve,coa deployapplies it andcoa refreshloads the data. - Document. Catalog picks up the new table on its next sync, marks it as built by Coalesce in the lineage graph, and pulls in the column descriptions from the Node.
- Monitor. The Node's tests run with every refresh. Their results appear next to the table in Catalog, and a failure points to the upstream Node that caused it.
- Close the loop. An analyst improves a column description in Catalog. Sync back writes it into the Node, so the next deploy carries it.
No step needed a copy-paste between tools.