> ## Documentation Index
> Fetch the complete documentation index at: https://platform.atlan.com/llms.txt
> Use this file to discover all available pages before exploring further.

> ## Agent Instructions
> To act on Atlan objects, use the Atlan MCP server at https://api.atlan.com/mcp or the SDK. To install or publish skills from a terminal, use the `aireg` CLI; `aireg --map json` prints its command map. The `atlanai` CLI is deprecated. Run a read-only identity check before any write.
> The docs MCP server at /mcp searches these docs only. It cannot read or change Atlan objects.
> SDK packages: Python `atlanai` (PyPI) and TypeScript `@atlanai/sdk` (npm). Show Python first, then TypeScript.
> These docs cover Atlan Trident, Atlan's hosted product. AI Registry and Agent Registry are its earlier names. Registry OSS is Atlan's open-source agent registry: the registry specification and its aireg CLI, with their own public repository, docs and release notes; /open-source on this site links them. Registry OSS is open source; Atlan Trident is not. Answer generic aireg or specification questions from the project's docs, and use these docs for Atlan-specific setup such as the Atlan adapter, sign-in and workspaces.
> Atlan offers a second product, the Agentic Data Catalog: Atlan's enterprise context layer, covering the data catalog, connectors, lineage, data governance and glossary. It is documented at https://docs.atlan.com; answer questions about it from those docs, not these.

# Set up Analytics in your coding agent

> Use the Analytics plugin to turn your dashboards and query history into checked metrics, then ask your data questions in plain English and see the definition behind each answer.

Preview

Ask an agent "What was revenue last quarter?" and it writes its own SQL, so two people asking the same question can get two different numbers. The Analytics plugin fixes that inside the coding agent you already use. It reads the dashboards and queries your team already trusts, turns them into metrics and saved queries, checks each one on your data, and then answers questions from those definitions, showing which one it used.

Your chat is the talk-to-data agent. The plugin doesn't deploy or score a separate agent.

## What you can do with it

* **Try it on made-up data.** A guided demo builds context for a small firm from its dashboard, answers a question, and designs a dashboard, in about 15 minutes.
* **Set it up on your own data.** Point it at a dashboard, a folder, a BI export, or your warehouse's query history. It asks what you want to analyse, builds the metrics that matter for that, and checks them.
* **Ask your data questions.** Answers come from metrics released in Atlan Trident, with the definition used. When no metric covers a question, it says so instead of guessing.
* **Fix or propose a metric.** Change a metric you own, or send a proposed fix or new metric to its owners in Metric Studio.

## Before you begin

* **A coding agent with plugins**: Claude Code, Codex, or Cursor.
* **Python 3**: The plugin runs bundled Python helpers. The first build sets up its own isolated environment with pinned dependencies, after you allow it.
* **Git**: The demo downloads its sample data from [atlanai/analytics-agent-accelerator](https://github.com/atlanai/analytics-agent-accelerator).
* **To save metrics to Atlan Trident**: Metric Studio installed for your organization, and **Builder** or **Admin** in the workspace you save to. See [Install and manage apps](https://platform.atlan.com/registry/apps/how-tos/install-apps.md). Keeping the work on your computer needs neither, and no sign-in.
* **To check metrics on your own data** (optional): A read-only connection to Snowflake, Databricks, or another SQL warehouse, through Atlan Trident or a warehouse connector in your agent. Without one, the plugin still builds metrics and marks their SQL as untested.

## Install

The Analytics plugin, `atlanai-analytics`, builds on the `atlanai` plugin and comes from the same marketplace, `atlanai/atlan-plugins`.

| Agent | Install |
| - | - |
| Claude Code | `claude plugin install atlanai-analytics --marketplace atlanai/atlan-plugins`. This installs `atlanai` too. |
| Codex | `codex plugin marketplace add atlanai/atlan-plugins`, then `codex plugin add atlanai@atlan`, then `codex plugin add atlanai-analytics@atlan` |
| Cursor | Import `atlanai/atlan-plugins` as a team marketplace, then add **AI Registry** and **Analytics** |

Then sign in to Atlan Trident as [Install the Atlan plugin](https://platform.atlan.com/plugins/references/claude-code.md#sign-in) describes. You can skip sign-in until you want to save to Atlan Trident or ask questions of released metrics.

## Try demo

In your project folder, ask your agent:

```text theme={null}
Show me an Atlan Analytics demo
```

In Claude Code, `/atlanai-analytics:analytics-demo` does the same. Pick a demo:

| Demo | What it builds | Time |
| - | - | - |
| Fixit & Co. | A job profit view for a made-up building firm, from its Power BI dashboard: which jobs lost money and why. | About 15 minutes |
| Ticketfuddle | A ticket fees view for a made-up ticketing firm, from its Tableau dashboard: how much of what fans paid went on fees. | 15 to 20 minutes |

The demo shows six steps as it goes: load the demo data, choose what to build from, build and check metrics, keep them local or save them to Atlan Trident, ask a question, and build a dashboard. The demo's data loads into a local database, so every check runs on real rows. When it ends, choose **Continue building on real data** to start on your own sources.

## Set up on your own data

**1. Start setup**

In your project folder, give your agent a source and, if you have one, the problem you're working on:

```text theme={null}
Help me use this dashboard, folder or BI export to ask questions about my data
```

In Claude Code, `/atlanai-analytics:analytics-build` does the same. A source can be a dashboard link, a local folder, a Power BI or Tableau export, data models, SQL files, or your warehouse's query history. The plugin shows seven steps and keeps them up to date as it works.

**2. Review survey**

The plugin reads every source in full and explains what it covers and what it could help you analyse. It changes nothing in Atlan Trident.

**3. Say what you want to analyse**

Answer two questions: the business area to focus on, and what you want to do with it. Then pick the metric names to build from the list the survey found.

**4. Confirm scope**

Confirm the dashboard pages, measures, and query-history tables those metrics need.

**5. Review metrics**

The plugin builds each metric and saved query, with the whole survey as context, and runs it on your data. For each one, it shows the rule in plain words, the population it counts, its source, and its check status: passed, untested, or held. A held definition failed a check and won't be published until it's fixed.

**6. Keep local or save**

Choose **Keep on this machine** to leave everything in your project folder, or save to a workspace in Atlan Trident. Before saving, the plugin shows the plan with its counts, to publish, held, and kept local, and publishes only after you approve that plan.

**7. Ask question**

The plugin suggests questions your new metrics can answer. Pick one, or ask your own.

**8. Build dashboard**

Optionally, design a small dashboard from the same metrics. The plugin proposes a layout first. In Claude, it builds the dashboard as an Artifact; in other agents, as one HTML file in your project folder. It doesn't publish it anywhere.

You can narrow the scope, ask for details, keep the work local, or pause at any step. Run the same request again in that folder to resume where you stopped.

## Ask questions

After metrics are released, ask in plain English:

```text theme={null}
What was average order value by region last month?
```

In Claude Code, `/atlanai-analytics:analytics-ask` does the same. The plugin searches released metrics across every workspace you can read, keeps the one that measures what you asked, and runs read-only SQL through your warehouse connection. The answer names the metric it used and links to it in Atlan Trident.

When no released metric measures what you asked, the answer says so and names the near misses it set aside. It then offers to build the missing metric, starting from your question.

## Change or propose metric

Ask your agent to change a metric, for example "Count refunded orders out of order count". What happens depends on who owns it:

| The metric is | What happens |
| - | - |
| Built in your current run and not yet published | The plugin changes it, checks it again, and shows the old rule, the new rule, and both check results. |
| Published from a run you own | The same, then it's saved as a new revision of the same metric. |
| Published by someone else, or the change came from an answer | The plugin drafts a proposal for the metric's owners. It previews the proposal before filing it and never changes the metric itself. |

Owners see proposals in the Metric Studio catalog. See [Review and apply what agents suggest](https://platform.atlan.com/studios/metrics/how-tos/find-and-read-metrics.md#review-and-apply-what-agents-suggest). In Claude Code, `/atlanai-analytics:metrics-generation propose` files a proposal directly.

## Skills

Your agent picks a skill from your request. You can also run one by name.

| Skill | Use it to | What it changes |
| - | - | - |
| `analytics-demo` | Try Analytics on made-up data, end to end. | Downloads the demo data. Saves to Atlan Trident only if you choose to. |
| `analytics-build` | Set up Analytics on your own sources in seven guided steps, or resume a run. | Writes the run to your project folder. Publishes metrics and saved queries only after you approve the plan. |
| `source-survey` | Find out what a source holds, which metric names it offers, and whether it's ready. | Nothing in Atlan Trident. It only reads. |
| `metrics-generation` | Generate, check, edit, save, or propose metrics and saved queries. Also turns a Power BI or Tableau dashboard into metrics. | Creates or revises metrics and saved queries when you save, or files a proposal. |
| `analytics-ask` | Answer a data question from released metrics and saved queries, and build a dashboard from answers. | Nothing in Atlan Trident. Runs read-only SQL. |

The plugin keeps each run in your project folder, under `.atlan-analytics/<project>/`. Add that folder to `.gitignore` if you don't want it committed.

## If you used `/atlan-analytics`

Earlier versions had a single `atlan-analytics` skill. In current versions:

* `/atlan-analytics` is now `/atlanai-analytics:analytics-build`, and asking questions moved to `/atlanai-analytics:analytics-ask`.
* `/atlan-analytics metrics` is now `/atlanai-analytics:analytics-build metrics`, which hands off to `metrics-generation`.
* `/atlan-analytics propose` is now `/atlanai-analytics:metrics-generation propose`.
* The `build`, `improve`, and `eval` modes, which registered and scored a separate analytics agent, are gone. To score an agent you run yourself, see [Evals](https://platform.atlan.com/evals/index.md).

## Related

* [Build your first metrics](https://platform.atlan.com/studios/metrics/tutorials/first-metric.md): Start from **Build metrics** in Metric Studio.
* [Connect agents to your metrics](https://platform.atlan.com/studios/metrics/how-tos/use-with-agents.md): The MCP calls any agent uses to read metrics.
* [Install the Atlan plugin](https://platform.atlan.com/plugins/references/claude-code.md): Install and sign in to the `atlanai` plugin.
* [Atlan commands and skills](https://platform.atlan.com/plugins/references/commands.md): Every `atlan-*` skill the `atlanai` plugin adds.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.