Agents can build your BI. Make sure they get the numbers right.
A semantic layer with everything under it. Define metrics once; every agent, app and report gets the same number. When a question is out of scope, the agent says so instead of making one up.
Text-to-SQL gives you confident wrong numbers.
An agent writing SQL against raw tables will get some answers confidently wrong, and you won't know which. In dbt Labs' 2026 benchmark, one frontier model got 84.1% of business questions right writing SQL against raw tables and 100% answering through a semantic layer.
Replaces Cube, dbt Semantic Layer, LookML and Raw-table text-to-SQL.
- Revenue, spend and every other metric defined once, in code
- Agents query metrics instead of guessing joins, so a failure becomes "I don't know" instead of a wrong number
- The same definitions behind apps, reports and every agent
- Freshness and quality checks agents read before they answer
How it works
- 01
Define metrics as code
Dimensions and measures live in your repo, reviewed like any other code.
- 02
Compile, don't guess
Queries compile to SQL deterministically, with checks on the data underneath.
- 03
Connect every agent
Claude, Cursor and your own agents connect through one MCP endpoint.
- 04
One answer everywhere
Apps, reports and agents read the same definitions. The numbers can't drift apart.
Enterprise, with our engineers
Our engineers model your core metrics (revenue, pipeline, spend, retention), test them against the reports you trust, and connect your agents to them through one gateway.
Self-serve, after launch
Define your metrics as code in the workspace, deploy, and connect your agents to the gateway. They query measures and dimensions; the platform writes the SQL.
Install from the marketplace
Everything below installs into your workspace as code. Use it as it is, or change anything.
KPI Scorecard
The company's KPIs on one page, defined once.
App
Ask Your Data
Questions in Slack, answered from governed metrics.
App
Model a new source
Draft semantic models for new tables, as a pull request.
Agent skill
Claude
Give Claude your company’s context.
Connector
Analytics workspace
A repo with pipeline, connections and a deployed app.
Starter
Questions
Why not let agents query the warehouse directly?
Benchmarks keep finding the same thing: agents writing SQL against raw tables get a meaningful share of answers wrong, while the same models answering through a semantic layer get nearly all of them right. dbt Labs measured 84.1% versus 100% in 2026.
How is this different from Cube or the dbt Semantic Layer?
Cube and the dbt Semantic Layer sit on top of a warehouse you already run. SegmentStream Cloud includes everything underneath it: collection, pipelines and the warehouse. The semantic layer isn't one more vendor on top of five.
Which agents can connect?
Claude, Cursor and any MCP client, through one endpoint with per-agent permissions and an audit trail.
Rather have it built for you? Start a pilot.
Our engineers model your core metrics (revenue, pipeline, spend, retention), test them against the reports you trust, and connect your agents to them through one gateway.
- One solution, on your data, in your own workspace
- First version live in two weeks
- Parity proven before anything is switched off
- Fixed scope and price, credited to your contract if you continue
Not ready for a pilot? Send the form anyway with your question. An engineer will reply, not a sales rep.
How the pilot worksYour company’s context, ready for every agent.
$ segmentstream workspace create acme --region eu
▸ workspace ready · data, objects, metrics and access rules in one place
▸ connect Claude, ChatGPT or Cursor with one connector
▸ tell your agent: “add our billing API as a source”
Our engineers build the first version on your data in two weeks, at a fixed price.
Or join the waitlist for self-serve:
Pilots are open now. Self-serve opens to the waitlist this fall.