Reference

Use this page to look up the main terms and access concepts used in the internal pilot.

Key terms

Term	Meaning
Culsma PAWS	The complete user-facing product, including AI workflows, web/app interfaces, workspaces, data services, and MCP connection.
Workspace	The top-level area that contains members, Lab Projects, samples, storage, and Analytics Projects.
Active workspace	The workspace currently used to authorize an AI or app request.
Lab Project	The operational project that contains laboratory samples and their context.
Analytics Project	A reusable analysis environment with separate datasets, roles, and lifecycle.
Dataset	A user-facing collection of related analytics tables or views.
Table	Structured analytics data organized into rows and columns.
View	A reusable, read-only result defined from analytics data.
MCP connection	The secure connection that allows an AI agent to use Culsma PAWS tools.
Change preview	A proposed write operation shown before it is applied.
Audit ID	An identifier used to trace an analytics query, load, or governed operation.

Workspace roles

Role	Main capability
Viewer	Read supported Workspace data and view public Lab Projects.
Editor	Viewer access plus supported routine create and edit actions.
Admin	Editor access plus supported administrative laboratory-data actions, including certain deletions.
Workspace Owner	Admin capabilities plus Workspace settings, invitations, role changes, member removal, and owner-only governance.

Only the Workspace Owner can manage Workspace membership in the current product. The admin role does not grant access to Workspace settings and does not allow the user to invite, change, or remove Workspace members.

Public Lab Projects can be viewed by Workspace members. Private Lab Projects require an explicit project role. Lab Project roles and Workspace roles are separate even when they use the same role name.

Lab Project roles and metadata

Role	Project metadata capability
Viewer	Read the project when its visibility and membership allow it.
Editor	Viewer access plus rename the Lab Project.
Project Admin	Editor capability plus supported project-membership management.
Project Owner	Admin capability plus change whether the Lab Project is public or private.

Lab Project metadata updates through the AI Agent are direct writes after explicit approval. A rename changes only the display name. Only the Project Owner can change visibility, and neither operation changes the project identifier, member roles, samples, or related Analytics Projects.

Analytics roles

Role	Main capability
Reader	Discover and query Analytics Project data.
Writer	Query data and add or update rows in existing datasets.
Publisher	Create or change schemas, tables, columns, and views.
Workspace Owner	Create Analytics Projects, update their names or descriptions, link Lab Projects, and manage analytics roles.

One person may have more than one analytics role. Linking a Lab Project does not copy its members or permissions into an Analytics Project.

MCP token permissions

Permission	Used for
lab.read	Read samples, projects, subjects, tags, and storage context.
lab.write	Preview and apply supported laboratory-data changes and imports.
lab.analytics.read	Discover and query authorized Analytics Project data.
lab.analytics.write	Create or maintain authorized analytics structures and data.

The token permission and the user’s Workspace or project role must both allow the operation. A token can limit the actions available to its owner, but it cannot grant a role or bypass missing Workspace, Lab Project, or Analytics Project access.

Pilot inputs and clients

• Sample intake accepts structured CSV or Excel files.
• Analytics table loading uses CSV data.
• The current Getting Started flow covers OpenAI Codex and Claude Code.
• Other MCP clients require organization-specific validation and setup.

This reference describes the current internal pilot. Availability may also depend on the features and permissions enabled for your organization.