Create an Analytics Project

For: Workspace Owners

Use this page to create a reusable Analytics Project and record its relationship to one or more Lab Projects. This workflow is performed entirely through your AI Agent and Culsma PAWS MCP.

Creating the project, linking Lab Projects, and granting roles are separate direct write operations. The AI Agent must show the exact plan and wait for your approval before each group of writes.

Before you start

You need:

the intended Culsma PAWS environment and active workspace;
an MCP connection with lab.analytics.write;
the Workspace Owner role;
a project name and clear purpose; and
the exact Lab Projects and initial collaborators, if any.

Choose a name that describes the sustained analysis purpose rather than one temporary question. The system creates or normalizes a stable project slug; check it before building datasets around it.

Start in your AI Agent

Replace the placeholders and copy this request:

In the active Culsma PAWS workspace, prepare a plan for an Analytics Project named [PROJECT NAME]. Its purpose is [PURPOSE]. Search for conflicting Analytics Projects, resolve the exact Lab Projects I name as sources, and show the proposed name, slug, description, project links, and initial role grants. Do not create, link, or grant anything until I explicitly approve the plan.

The AI Agent should:

state the selected environment and authenticated workspace;
confirm that you are the Workspace Owner and have analytics write access;
search visible Analytics Projects for a conflicting name or slug;
resolve every requested Lab Project in the same workspace;
show the proposed project identity, description, links, and grants; and
stop before the first write.

Review and approve the plan

Plan item	What to check
Workspace	Correct active workspace and environment.
Name and slug	Clear, stable, and not already used in the workspace.
Description	Explains the analysis purpose and what belongs in the project.
Lab Project links	Exact source-project identifiers; no unrelated project.
Initial collaborators	Existing workspace members only.
Roles	Reader, Writer, and Publisher grants match each person’s task.

Unlike sample change sets, Analytics Project creation is not staged. After you approve, the AI Agent calls the direct MCP create operation.

Use an approval that limits the first action:

I approve creating the Analytics Project exactly as shown. Create only the project first, then fetch it and stop before linking Lab Projects or granting roles.

This separation lets you verify the server-generated Analytics Project identifier before dependent operations use it.

Update the project name or description

A Workspace Owner can update an active Analytics Project’s display name or description through the AI Agent. The stable slug and project identifier do not change.

Ask for a read-back before approving the direct write:

Fetch Analytics Project [PROJECT ID]. Show its current name, description, and stable slug, then propose changing [NAME OR DESCRIPTION] to [NEW VALUE]. Do not update it until I approve the exact project and final value.

After approval, the AI Agent updates only the named fields and fetches the project again. If you intend to remove the description, say explicitly that it should be cleared; do not rely on an omitted or blank field. Renaming does not rename schemas, tables, views, linked Lab Projects, or the stable slug.

Link Lab Projects

After project creation and read-back:

confirm the canonical Analytics Project identifier;
review each exact Lab Project identifier;
approve the listed links; and
ask the AI Agent to list the links again after creation.

Linking records provenance and scope. It does not copy Lab Project data, create warehouse tables, or transfer Lab Project permissions.

Grant initial access

Grant roles only after the Analytics Project identity has been verified. The grantee must already be a member of the active workspace.

Give Reader to someone who only needs discovery and read-only queries.
Give Writer to someone who must query and load or change rows.
Give Publisher to someone who creates or changes schemas, tables, columns, or views.

A Publisher who also needs to query or load data may need an additional Reader or Writer grant.

Review the evidence

A completed setup should return:

environment and workspace;
Analytics Project ID, name, slug, description, and active status;
created-by and created-at information;
the exact linked Lab Projects; and
the active grants created for the project.

Ask the AI Agent to fetch the project, list its Lab Project links, and list its active role grants. This MCP read-back replaces a Web/App confirmation screen.

More requests to try

Create only the approved Analytics Project. Do not link sources or grant roles yet. Fetch it after creation and show its canonical identifier.

Rename Analytics Project [PROJECT ID] to [NEW NAME]. Keep its slug, description, links, grants, schemas, and data unchanged. Show the current and proposed names and wait for approval before updating it.

Link Lab Projects Alpha and Beta to Analytics Project [PROJECT ID]. Show the resolved Lab Project IDs and wait for approval before writing.

List the active roles for Analytics Project [PROJECT ID] and recommend the minimum grants needed for one Publisher who will also load data.

If setup cannot continue

What you see	What to do
The name or slug already exists	Reuse the existing project if it is correct, or choose a different stable identity.
You are not the Workspace Owner	Ask the owner to create the project or perform the owner-only operation.
Analytics write scope is missing	Reconnect with the approved `lab.analytics.write` access; do not broaden unrelated scopes.
A Lab Project is not in the workspace	Stop and resolve the correct source project; cross-workspace linking is not allowed.
A collaborator is not a workspace member	Add or invite the person to the workspace before granting analytics access.
Project creation succeeds but a later link fails	Keep the verified project, resolve the exact failed link, and retry only that operation.
Read-back does not match the plan	Stop before creating datasets and report the canonical project ID plus the mismatch.

You have completed setup when the project identity, linked Lab Projects, and active grants are all verified through MCP read-back.

Next, continue to Build an Analysis Dataset.