Data Connections
Connect the sources your team already uses without pasting credentials into chat.
Create encrypted workspace connections through guided forms or OAuth, let your AI inspect safe samples, then keep dashboards and pipelines attached to their source and refresh path.
Workflow
How the feature works
Choose the source
Use a file, database, SaaS connection, API, Google source, GitHub, or MCP server.
Authorize securely
Open the DataUnmess connection flow. Credentials are never accepted inline by the AI.
Inspect before building
Read schemas, fields, and a small sample before deciding the dashboard or pipeline shape.
Keep the refresh path
Save query or pipeline metadata so Refresh returns to the upstream source.
Supported sources
One workspace for files, systems, databases, and APIs
The connection catalog keeps source identity and authorization outside the AI conversation while making each source available to dashboards, analysis, and pipelines.
CSV
Excel
JSON
Parquet
Google Sheets
GA4
HubSpot
PostgreSQL
MySQL
ClickHouse
REST API
GitHub
MCP serverAvailable now
Files, operational systems, databases, APIs, and agent tools
DataUnmess supports direct database queries, live Google Sheets reads, managed SaaS imports, REST API connections, GitHub access, and third-party MCP servers. CSV, Excel, JSON, and Parquet files can be imported into the workspace.
Connection strategy
Query live or materialize a managed dataset
Use a direct query when the source can answer the chart or analysis efficiently. Create a Data Pipeline when the source needs cleaning, joins, reusable preparation, scheduled imports, or a managed destination.
Google Sheets dashboards can remain connection-backed. GA4 and HubSpot use managed import bundles that seed recent data, create datasets, schedule daily repair-window refreshes, and build an initial dashboard.
Security
Keep authorization in the connection flow
An MCP-connected AI can request a secure deep link for a new connection, but it should never ask the user to paste credentials into chat. Database passwords, API tokens, and private app tokens stay in the DataUnmess connection form.
Use read-only or least-privilege credentials whenever the source supports them.
Example prompts
Ask for the outcome in business language
Connect my PostgreSQL database, inspect the available schemas, and suggest three useful dashboards.
Read this Google Sheet and show me the columns and ten sample rows before building anything.
Import my HubSpot CRM, create the standard dashboard, and schedule the daily refresh.
FAQ
Data Connections questions
Can I paste database credentials into my AI client?
No. Ask the AI to create a connection. DataUnmess returns a secure deep link where credentials are entered and encrypted.
Do all sources need to be copied into DataUnmess?
No. Supported sources can be queried directly. Materialize a managed dataset when you need reusable cleaning, scheduled imports, or a stable destination.
How do dashboards stay connected to a source?
Saved cards retain dataset query metadata or connection query metadata. Managed dataset cards refresh through their upstream pipeline first.
Continue exploring