datasetslisted

# CloudSprite datasets (read) Find and describe datasets in the signed-in user's current project. Execute **MCP read tools** — do not generate scripts, do not call the REST API with a token, and do not modify parameters, tags, or notebooks. ## Setup 1. Confirm MCP is connected. If not, tell the user to finish OAuth. 2. `get_scope` (or `whoami`). If team/project is unset, `list_teams` / `list_projects` and `set_scope` with their choice. 3. Then search. ## Finding datasets Map natural language to `search_datasets`. Typical filters: - Name glob or substring - Parameter predicates (values are strings — `lot=15` means `"15"`, not `15`) - Tags - Source / file type when the tool supports it Examples: | User says | Do | |-|-| | "Which datasets in this project have lot=15?" | `search_datasets` with parameter `lot` = `"15"` | | "Show datasets tagged needs-review" | `search_datasets` by tag | | "What is on dataset ABC?" | `search_datasets` to get the id, then `get_dataset` | | "What traces does it have?" | `get_dataset` (inventory) then `get_trace_summary` per trace | | "Which notebooks include these?" | `list_notebooks` / `get_notebook` | Always name the **project** you searched and how many rows matched. If the list is long, summarize and offer to narrow. ## One dataset `get_dataset` returns parameters, tags, trace inventory, and provenance (source type, producing run). Use it before answering "what's on this file." For numeric shape (x-range, point count, min/max/mean,