Your Context Factory, live in a week
A SignalPilot engineer or a certified partner works on your schema, under NDA. We ingest your BI tools, dashboards, dbt project, and agent traces for you, so your team starts from a strong draft, not a blank page. By Friday the evals are running and the sandboxed cloud agent is live for your first team.
- 1MonScope
- 2TueIngest everything
- 3WedCurate the draft
- 4ThuEvals and gate
- 5FriShip
- parsed from
- dbt
- semantic layer
- dashboards
- agent traces
- whole estate · first team chosen
- 62 definitions drafted · 3 PRs opened
- 12 open questions settled · named
- 118 graded questions · gate on
- agent live · eval coverage 91%
Two ways in
Guided, or self-serve.
The integration week
Included on annual Scale and Enterprise plans, quoted per engineer-week on Team, or delivered by a certified partner.
See pricing →The four-step quickstart
Control plane, warehouse, plugin, MCP gateway, on the open-source stack. Under an hour if your dbt project is clean.
Read the quickstart →What we need from you
Five things, one hour.
Everything runs under NDA. Queries run on your warehouse's permissions and are logged. Bring your own Anthropic, Bedrock, or Vertex key if you prefer.
- Read-only warehouse credentials, or a warehouse schema.
- Read access to the dbt repo through the GitHub app, with permission to open pull requests.
- The dashboards your leadership trusts and the queries behind them.
- The first team the agent goes to, and the person who owns its questions.
- One hour with your data lead on day one, and curation time on day three.
A typical week
Scope on Monday. Ship on Friday.
Most teams go live in a week. Messier dbt projects take a little longer.
- 01
Day 1 · Scope
Pick the first team the agent goes to. Walk the schema, dbt models, and dashboards together. Agree on what done looks like.
- 02
Day 2 · Ingest everything
We ingest your BI tools, dashboards, dbt project, and agent traces. Our agents trace how your numbers are calculated, draft definitions, flag conflicts, and open pull requests where the dbt project needs fixing.
- 03
Day 3 · Curate the draft
Your team settles the open questions we found, like two dashboards with different definitions of revenue, and names each variant with its owner.
- 04
Day 4 · Evals and gate
Graded questions with failure cases, the nightly run wired, the pull request check turned on.
- 05
Day 5 · Ship
The sandboxed cloud agent goes live for your first team, from Claude Cowork, Codex, or Slack. You get the coverage number and the eval report.
What you have at the end
- Named definitions in dbt, Apache Ossie, and Markdown, in your repo.
- A passing eval suite for your first team's questions, with the failure cases covered.
- The nightly run and the pull request gate, on.
- The sandboxed cloud agent live for your first team, with an evidence graph on every answer.
- A coverage report: models covered, models to do next.
Brightwater Goods · Fri 16:00
- ✓Graded questions118 of 118 passed
- ✓Failure cases caughtfan-out join · refunds left in · wrong window
- ✓Definitions curated62 · 12 open questions settled · Markdown + Apache Ossie
- ✓Eval coverage91% of models
- ✓Pull request gateon · affected evals rerun on every PR
- ✓Agent liveClaude Cowork · Codex · Slack
- mart_wholesale_margin
- mart_weekly_pipeline
- mart_cohort_retention
- +2 more
After the week
Run it yourself, or hand us the loop.
Your team owns the loop. Thumbs-downs and gaps arrive daily as recommendations, usually as pull requests that add the missing evals or context.
A SignalPilot engineer runs the weekly review and stands behind 90% precision. See Managed →
Roll the agent out to the next team. With your Context Factory in place it takes days, not another week.
FAQ
Before you book the week.
Our dbt project is a mess. Does that stop us?
No. Fixing the project is part of the week. We open the pull requests; you review them.
Do we have to replace our BI tool?
No. Definitions ship in the Apache Ossie format so your BI tool can read them too.
Can a partner run the week instead?
Yes. Certified partners deliver the same week on the same tooling.
How much data leaves our warehouse?
Query results reach the sandbox for the answer and its evidence. Definitions live in your repo.
Start the week.
Bring your dbt project. We ingest the rest.
Open source · Apache 2.0 · GitHub · Slack