Home
New to AI
All 47 tools →
Features
Use cases
Live demo
Impact analytics
Pricing
Blog
FAQ
About
CLAUDE.md Generator
Few-Shot Example Set Builder
MCP Config Wizard
System Prompt Generator
Prompt Optimizer
RAG Chunking & Retrieval Planner
ReAct Prompt Converter
Skill File Generator
Structured Output Schema Builder
Subagent & Slash Command Generator
Tool / Function Schema Generator
Workflow to Agent Blueprint
Free tool
Golden Dataset Manager
Curate, version, and review your test cases as a first-class asset — the substrate the Eval Runner, Prompt Regression Diff, and Blast Radius all run against.
Get started-
Every case tracks provenance (synthetic, production sample, or manually authored) — with a loud warning when a dataset is entirely model-generated, the most common quiet failure mode in eval sets.
-
Nothing is ever overwritten: every change is a new version with a diff against the last, because historical run results stay bound to the exact version that produced them.
-
A one-click "capture a failing case" turns a real production failure into a test case in seconds, plus a coverage view that names the gaps in your tags and difficulty spread.
Frequently asked questions
None. Golden Dataset Manager never calls a model — it is a pure curation, versioning, and review surface for test cases you write, import, or promote from AI-generated sources like the Eval Set Generator.
They are the same underlying asset. A golden dataset is simply an eval set that has been curated over time — versioned, tagged with provenance, reviewed by a teammate, and checked for coverage gaps — rather than a one-off generated document.
Yes. Curating, versioning, and reviewing datasets writes to your organization's Asset Library and requires a Pro or Team plan; a Free account can still view and run any eval set it already has.
Yes, directly and immediately — both already select from any saved eval_set asset in your library, and a golden dataset is exactly that, just with richer curation on top.
They are never deleted. Every version stays available to view, download, and diff against — so a past Eval Runner or Regression Diff result always stays bound to the exact dataset version it actually ran against.