I'd score AQL a 9 or 10. It's up there with the best tools I've used. It's really cool that we can now define metrics based on other metrics, stacking them on top of each other.
Ian
BI Engineer, DNSFilter
Build your BI dashboards
with coding agents
Use coding agents to build models, metrics, and dashboards. Locally on your computer with Claude, Codex and Cursor. Or in the browser with our Development Copilot.
secure.holistics.io/reports/revenue-dashboard
Live preview
Your dashboard will build here
Revenue dashboard
Live claude · ~/analytics
>
⠋ Build Revenue dashboard
└ ✓
holistics sync-code · live
branch: feature/revenue-dashboard
› esc to interrupt · ctrl+t to hide tasks
Two ways to build with AI
In the workspace · Development Copilot
Your agentic BI developer, built into Holistics
Build datasets, define metrics, and generate dashboards in plain English, right in your browser. No setup required.
Build datasets in plain English
Create and modify datasets from a description. The Copilot auto-suggests relationships and defines metrics, dimensions, drill-downs, and underlying data.
Generate whole dashboards
Spin up multiple visualizations, filters, and controls and arrange them into a cohesive layout, from a single prompt.
Document the semantic layer in bulk
Auto-generate descriptions for model fields, datasets, and tags from existing names, labels, and logic. Your layer stays self-explaining without the manual writeup.
Write commit & PR descriptions
Generate commit messages and pull-request descriptions from your current changes and your PR template, ready for review in one click.
In your IDE · Local agentic development
Bring your own coding agent
Connect your favorite coding agent, Claude Code, Cursor, or Codex, to your BI codebase. Run queries, build metrics, and preview results live through Holistics' CLI and MCP tools.
A real session, a coding agent building Holistics analytics from the terminal.
1
claude · ~/analytics
› /setup-amql-development
Holistics CLI authenticated
MCP server connected
Live sync started
Watching 12 models · 4 dashboards
› ready — start prompting
Clone & connect
git clone your repo, then run /setup-amql-development. One command wires up the CLI, MCP, and live sync.
2
revenue.page.aml
Add a Revenue dashboard with QTD & YTD KPIs
+ Dashboard revenue {
+ KPI qtd_revenue { … }
+ KPI ytd_revenue { … }
+ BarChart monthly_rev { … }
+ }
Build by prompting your agent
Ask Claude Code, Cursor, or Codex. It explores your workspace over MCP and writes the AML.
3
app.holistics.io/revenue
QTD$4.8M
YTD$14.2M
Margin67%
Preview live, instantly
Every save syncs to a cloud dev branch in seconds. See the real dashboard and tweak it back.
4
Add Revenue dashboard #42
+128 −4 · 3 files changed
revenue.page.aml · orders.model.aml · users.model.aml
AML validate passed
1 review approved
Merge pull request
Ship like software
Commit, open a pull request for review, and merge. Or publish straight to production.
How they connect together
Your agent edits AML on disk. Three connections do the rest: code, MCP, and Git.
MCP — read & reason
Your agent inspects schema, runs AQL, and searches the docs against your real warehouse, so it's grounded in what's actually there instead of a guess.
CLI sync-code — live preview
Every save streams to a cloud dev branch; the real dashboard renders within seconds, and edits in the UI sync back to your files.
Git — ship
Clone and push your BI as code. Open a pull request, validate in CI, and merge to publish the prod branch.
Why coding agents can build your BI analytics
Agents need a substrate they can read, reason on, and change safely. Here's what Holistics gives them.
Analytics-as-code
Every model, metric, and dashboard is typed AML in Git. Agents edit real code, with history, branches, and pull requests, instead of clicking through a config UI.
A semantic layer agents reason on
Agents generate AQL against your governed definitions instead of raw text-to-SQL, so the query is grounded in metrics your team already trusts and you can read what it actually did.
Reachable over MCP
The MCP server gives any agent live workspace access. It can explore models, run queries, and inspect schemas, reasoning from the same definitions everywhere.
Whatever builds it, the output is governed
Agent-authored or hand-written, every change runs through the same guardrails.
Grounded in the semantic layer
Agents compose from endorsed metrics and relationships instead of inventing numbers from scratch.
Versioned and reviewed
Every change is code in Git: diffs, pull requests, and approvals before anything reaches production.
Traceable lineage
Each metric traces back to its definition and the models it's built from, so answers hold up.
What teams building on Holistics say
Bulk edits and refactors happen in code, not by clicking through a UI for an afternoon. It keeps a lean data team fast.
Senior data engineer
Funding Societies
Non-technical users run their own analyses now. The data team stopped being the queue.
Data team lead
Optimal Workshop
Frequently asked questions
Which coding agents are supported?
Claude Code, Cursor, GitHub Copilot, and Codex all work with the local workflow via the Holistics MCP server and CLI. The in-product Development Copilot needs no setup at all.
Do I need to set anything up for the cloud Copilot?
No. The Development Copilot is built into the Holistics workspace. Open a dataset or dashboard and start asking. Local agentic development requires connecting your repo and running /setup-amql-development.
Is agent-generated work governed and safe?
Yes. Whatever creates a change, it's typed AML grounded in your semantic layer, versioned in Git, and reviewable through pull requests before it reaches production.
How do changes ship to production?
Through the same path as code: commit and push, open a PR, review and merge, then publish, either manually, via the Publish API, or with an auto-publish GitHub Action on merge.
See it on your data
A 30-minute walkthrough with a Holistics engineer. We'll connect a sample of your warehouse and build something with an agent, live.