Misata Logo
Misata Studio

What changed

A plain record of what has shipped, newest first. Something missing that you need? Tell us.

Last reviewed November 2026

November 2026

Agents that use the API

  • A guide written for coding agents: which route to use, the order of calls, how to wait for a build, report what held, take the data, check numbers against the answer key and change a dataset in words. Each deployment serves it as plain Markdown.
  • Studio's API page hands you the block to paste into your agent's instructions.
  • Keys can be scoped: read only, a monthly credit cap, and an expiry. An agent is told what its key may do.
  • Load a dataset into HubSpot or Salesforce from the API: plan first, then apply, with records upserted on their Misata id.
November 2026

Editing after the build

  • Each version says what it was made from: the lever you moved and its old and new value, a redrawn curve, a correction or a new draw.
  • A number you set in Adjust is shown as yours, beside the request that first stated it.
  • Adjust shows one lever per cause, named by the table you download, and offers Apply only where something changed.
October 2026

Data that loads into Salesforce and HubSpot

  • Export a dataset as Salesforce's standard objects or HubSpot's objects: one CSV per object in load order, every link named by an external id.
  • Each export carries a loader script that creates the custom fields or properties, checks picklists and pipelines against your org or portal, and upserts in order.
  • Import a Salesforce object describe or HubSpot properties as a schema, custom objects included, and build data for that shape.
  • The SaaS and sales shape reads as a CRM: prospects beside customers, each a business of its own country, a buying committee at each, and won, lost and open deals, the won ones booking the contracts.
October 2026

Pepperwind, a whole company as data

  • One company across 223 tables: storefront, billing, support, marketing, product and the ledger, joined by the same customers.
  • A free preview on Hugging Face under CC BY 4.0, and a page that runs SQL on it in your browser.
  • Graded questions that name the misreading a wrong answer equals.
October 2026

Answer keys for evals

  • Every dataset carries its answer key: each standard number with its SQL and its value by month, as JSON and CSV.
  • Questions drawn from the key with their answers, and the usual wrong readings of each number worked out on the rows.
  • Paste a dashboard's numbers or its SQL into the Story tab and see whether they match, and what a wrong one computed instead.
  • Exports for ClickHouse and Postgres that load in one step with every foreign key.
October 2026

Your own schema, with your numbers

  • Import a schema from SQL DDL, Prisma, dbt, Django, SQLAlchemy, Rails, Mermaid, OpenAPI, JSON Schema, BigQuery, Avro, DBML or CSV.
  • State totals, rates, shares and balances on your own tables. The plan shows each one before you build, and the build holds it.
  • A number that cannot hold with the others is named before any row is made, with what to change.
  • The designer gives each column a kind with its own editor, and shows sample values on every card.
October 2026

A certificate that says what was not held

  • Each stated number is listed with the months it covers and how closely it was met.
  • A number the build could not hold reads not met, with the reason in one sentence.
  • The design's own rules are checked on every build: keys, lifecycles, dates in order, and rows inside their parent's life.
October 2026

Builds and edits you can come back to

  • Reload or leave a build and come back to it where it was, with every step it has taken.
  • Corrections in words, redrawn curves and exact rebuilds run like builds, with progress, cancel and resume.
  • An API for apps: send a prompt, take the data, revise it in words, and get a signed webhook when each version is ready.
October 2026

Curves you can draw on real columns

  • Pin a curve to a column, such as order totals, or to how many rows land in each month.
  • Before you build, see what the curve means for the rows: how many there are each month and what each one is worth.
  • Redraw the curve on a finished dataset. It shows the cost first, and nothing is charged when the curve is unchanged.
  • The finished dataset compares every month you asked for with what was generated, in your currency.
October 2026

Builds you can leave

  • A build keeps going when you close the tab and tells you when it is done.
  • Builds survive a server restart: they are picked up and resumed.
  • Credit packs of 50, 200 and 900 credits for $9, $29 and $99.
October 2026

Realism, checked and shown

  • Every certificate carries the realism scorecard, and exports include it.
  • Planted errors can gather around one thing, such as a single shop or a single month.
  • Fixes for growth ramps, lengths, late follow-ups, payslips after someone has left, and rows without names.
  • Batch Release World: a ready-made pharmaceutical batch release dataset with its audit and docs.
September 2026

Realism you can rely on

  • Dates are checked across tables: a visit can no longer come before the patient registered, or a payment before its visit.
  • An expert-style review reads the finished data, flags what is implausible and rebuilds once, keeping the result only if it still verifies.
  • Believable magnitudes: fees, salaries and prices are set for the business and the region, not left at defaults.
  • A schema you paste is followed exactly and any difference is reported; a vague request is designed on purpose and lists what was assumed.
  • Datasets get short, readable names.
September 2026

A workspace that fits your role

  • Tell Studio who you are once and it reshapes the home page, the examples and the export order.
  • Business metrics such as revenue over the years are worked out from your data, with the query behind each.
  • Draw the curve you want and the data follows it exactly.
September 2026

Design it like a canvas

  • Tables on an ER diagram, with kits, formulas, ranges, auto-layout, undo and redo.
  • Explore in your browser: charts, dashboard, insights, profiles and your own SQL.
  • Twenty export formats, and SQL for seven databases.
September 2026

Worlds with no data behind them

  • Invented and future worlds are written with cited research; sources are shown on the dataset.
  • Names, cities, phones and currencies agree with each other for any place, real or invented.
  • Anti-money-laundering data with labelled typologies, alerts and cases as ground truth.
September 2026

Studio first

  • One focused home page, use-case guides, honest comparisons with dates, and pages that explain how we verify and how your data is handled.