Find the right DataLens dashboard and dataset from chat
Connect Yandex DataLens to the AI agent: collections, workbooks, dashboards, datasets and connections. The agent locates a report and shows what already exists.
How the agent works with Яндекс DataLens
The agent calls api.datalens.tech and walks the account through RPC endpoints: it lists the directory of collections and workbooks page by page, descends into them and collects dashboards, charts, datasets and source connections. It then maps out how the objects relate (which chart sits on which dataset, which dataset on which connection), spots duplicates of the same report and points you to the one you need.
The agent reads report configuration and can retrieve dataset rows through the experimental getDatasetData endpoint in API v3. It uses field GUIDs, checks ordering for paginated exports, and marks incomplete results.
The API supports object creation, updates, deletion, access management, and workbook export/import. Changes follow the user request and granted permissions, and the agent checks the result. A workbook moves between collections by export and import, and the agent waits for the file to be ready.
The connection is a Yandex Cloud service-account authorized key: the connection finds the DataLens organization from the key itself, and the agent renews the IAM token without anyone's help. The walk through the account is paginated, so on a large organization the agent takes the directory in parts and builds up the object links as it descends rather than in one call. It holds the request rate within the limit and names an empty result plainly, telling «no reports» apart from «reports not visible» instead of passing one off as the other.
Integration scenarios
Navigating collections and workbooks
The agent walks through your organization's collections and workbooks and points you to the dashboard, dataset or source connection you need.
Report inventory
The agent lists dashboards, charts and datasets, spots duplicates of the same report and shows how the objects relate to each other.
Access and change audit
The agent shows permissions on objects, checks what a specific user can reach and pulls change history from any date you name.
Moving a workbook between collections
The agent exports a workbook, waits for the file to be ready and imports it into another collection in the same organization.
Licenses and embed links
The agent pulls the organization licenses and the embed links of objects and shows which reports are exposed outside and what the current plan still covers.
What a dataset is built from
The agent opens up the dataset fields, sources and settings so you can see which breakdowns exist at all; for the numbers themselves it goes to the source system.
Яндекс DataLens examples
В коллекции «Финансы», воркбук «Отчётность 2026», и у него есть близнец.
Обошёл 6 коллекций и 23 воркбука: второй такой же дашборд лежит в «Архиве» и тянет тот же датасет. Доступ к обоим только у пяти человек из отдела.
How to connect
Before you connect
- A Yandex Cloud service account with a DataLens role in the right organization.
- Viewer access for the service account on collections and workbooks.
- DataLens admin rights for whoever grants the role.
- 1
Create a service account in IAM
Yandex Cloud console → “Identity and Access Management” → “Create service account”, in a cloud of the organization where DataLens runs.
- 2
Grant it a DataLens role in the org
Cloud Center → organization “Access bindings” → “Assign bindings” → the service account: datalens.admin for full access or datalens.visitor to sign in.
- 3
Assign the viewer role on collections
An organization role alone does not expose content: assign datalens.collections.viewer and datalens.workbooks.viewer to the account on each collection and workbook.
- 4
Download the JSON key from the account
“Create new key” → “Create authorized key” → “Download file with keys”, then paste the whole JSON. The connection finds the organization by itself.
FAQ
With a Yandex Cloud service-account key: paste the whole authorized-key JSON file, and the connection finds the organization by itself. DataLens grants permissions at the organization level, so the account needs a role there, plus viewer access on collections and workbooks, otherwise their contents stay invisible.
The structure of your account: collections, workbooks, dashboards, charts, datasets and connections, how they relate, their permissions, change history, organization licenses and embed links.
Yes, through the experimental getDatasetData endpoint in API v3, when the source and granted permissions allow it. Dashboard configuration itself does not contain the figures.
The agent can create, update, and delete objects, manage access, and transfer workbooks as requested and within the granted permissions.