Analytics Projects Overview

An Analytics Project is a reusable analysis environment with its own datasets, collaborators, permissions, and lifecycle. Use one when data needs to be structured, queried repeatedly, shared with a defined group, or maintained independently from everyday laboratory records.

AI Agent only in the current pilot

Analytics Projects currently have no Culsma PAWS Web/App interface. Every analytics task begins in your AI Agent and is carried out through the Culsma PAWS MCP connection.

This changes how you review the work:

Everyday Lab Work	Analytics Projects
Web/App can often provide a visual alternative or confirmation.	There is no analytics Web/App alternative in the current pilot.
Sample and storage writes normally use a staged change set.	Most analytics writes are direct MCP operations after an explicit execution plan and approval.
You may verify a saved record visually.	You verify schemas, tables, columns, row counts, SQL, status, and audit identifiers through MCP evidence.

For a read-only query, the AI Agent must discover accessible data, dry-run the SQL, execute only in read mode, and return reproducible evidence. For a write, the AI Agent must show the exact target and operation, stop for your approval, execute only the approved MCP action, and then read the result back.

Lab Project or Analytics Project?

Use an immediate project-data question when…	Use an Analytics Project when…
You need one read-only answer from data already available for analysis.	You need a reusable analysis dataset.
No new table or maintained data is needed.	Data will be loaded, structured, or maintained over time.
One established source contains the required information.	Several Lab Projects or external files may contribute data.
No separate analytics roles are needed.	Readers, Writers, and Publishers need different access.

Linking a Lab Project records its relationship to the Analytics Project. It does not copy data, create a dataset, or grant Analytics Project access to Lab Project members.

Analytics roles

• Reader: discovers available schemas, tables, and columns and runs audited read-only queries.
• Writer: reads data and adds, updates, or loads rows in existing tables.
• Publisher: creates and changes schemas, tables, columns, and reusable read-only views.
• Workspace Owner: creates Analytics Projects, links Lab Projects, and manages analytics role grants.

A Publisher role does not automatically include Reader or Writer behavior. A person who creates a structure and also needs to query or load it may need an additional Reader or Writer grant. Use the smallest set of roles that supports the work.

Access requirements

Analytics access has two independent parts:

1. the Culsma PAWS MCP connection needs lab.analytics.read for read-only work or lab.analytics.write for changes; and
2. the user needs the appropriate active role for the selected Analytics Project.

A write-capable MCP token does not replace the required project role. If either part is missing, ask the Workspace Owner or pilot coordinator to update the minimum necessary access.

Choose a task

What you want to do	Start here
Create an Analytics Project and link its source Lab Projects	Create an Analytics Project
Create a schema, table, or reusable read-only view	Build an Analysis Dataset
Discover data or answer a read-only question	Explore and Analyze Data
Append or replace CSV data or change a table structure	Maintain Analytics Data
Grant or revoke Reader, Writer, or Publisher access	Share and Govern an Analytics Project

Evidence replaces visual confirmation

Do not treat an analytics task as complete because the AI Agent says it succeeded. Review the evidence appropriate to the operation:

• selected environment, workspace, and Analytics Project identifier;
• linked Lab Projects when relevant;
• schema, table, view, and column names;
• exact SQL for read-only analysis;
• row count, truncation status, load mode, and operation status;
• query, DDL, or load audit identifier when returned; and
• a read-back listing or verification query after every write.

Typical path

1. A Workspace Owner creates the Analytics Project and links relevant Lab Projects.
2. The owner grants the required Reader, Writer, and Publisher roles.
3. A Publisher creates the schema and reusable table structure.
4. A Writer loads and maintains rows.
5. Readers and Writers run audited, reproducible queries.
6. Every write is verified through MCP read-back evidence.

Start with Create an Analytics Project if no project exists yet. If the project and dataset already exist, continue to Explore and Analyze Data.