Solo: everything in your own AI
Solo is the way of working in which all the thinking happens in your own AI tool, on your own subscription. Plainwater AI supplies the tools, the data access, the rules and the storage, and incurs no model costs of its own. It is an extra choice next to the regular connection, not a replacement.
Why Solo exists
A Claude Max or Pro subscription delivers far more agent work for the same money than API credit. Anyone who already has such a subscription does not want to pay model costs in the platform on top. With the regular connection the work falls back on platform AI in a few places: subtasks, connectors, saving dashboards and datasets, and the automatic starters on a work item. In Solo all of that happens in your AI.
What differs from the regular connection
| Topic | Regular connection | Solo |
|---|---|---|
| Subtasks by platform agents | Yes | No: the tool does not exist in Solo |
| Calling connectors | Via handover | Directly |
| Saving dashboards and datasets | Via a platform conversation | Directly, as a draft |
| Automatic starters on the work item | Start platform AI | Start nothing |
| Handing work over | Yes | Yes, if you explicitly want to |
| Model costs in the platform | On handover and subtasks | None |
The difference shows in the tool list: in Solo your AI does not even see the tool for starting subtasks. That is why
Solo is a connection of its own with its own address (/mcp/solo) and its own token; see
connecting your own AI tool.
What stays the same
- Data access per source, pseudonymisation and the cohort threshold of ten.
- Your permissions, read again on every call.
- The audit: every tool call is on the work item.
- Local sessions must be switched on for the customer.
How a Solo session runs
- You connect
plainwater-soloin your AI tool and sign in. - Your AI starts a session on a work item.
- It sets the work item's sources and chooses one.
- It works with the tools: reading metadata, measuring quality, creating a disposable copy, running SQL if the source allows it.
- It calls connectors, for example to read a Power BI model.
- It saves a dataset or dashboard draft.
- A person publishes the draft in the web app.
- If you do want the platform to continue, you hand the work over.
Solo suits work you steer yourself and whose costs you want to keep predictable. For work that should keep going while your laptop is closed, or where you want several agents working in parallel, the regular connection or the web app is handier.
Example
An analyst with a Claude Max subscription is asked to build a dashboard of outstanding receivables. She connects Solo in Claude Code, starts a session on the work item and picks the source finance-prod. Claude Code reads the metadata, writes a query, saves it as a dataset draft and builds a dashboard draft on it. She publishes both in the web app. From then on the dashboard refreshes on schedule, and not a single model was involved in the platform.