For API teams about to let a few customers in early
Beta CharterA beta label is a contract with fewer guarantees - write it down before the first account enrolls.
Describe the feature, paste the stage contract, the accounts you want in and any beta that never ended. Your browser ranks the program shape, reads the eleven contract rows of every stage, checks the stage is on the wire, the time box, freeze and extension rule, holds every account to the entry criteria and the capacity cap, reads the sign-offs, and decides every perpetual beta: GA as-is, next version or kill. Then the charter, and the rescue.
The feature, and a short interview free
The contract, the cohort and the live betas free
The checks free
What the page flags
The gates
The shapes, ranked
The stage contract, row by row
The cohort against the entry criteria
Betas that never ended
Area by area
Your contract table free
The table the skill drafts first - every row for every stage - with the empty rows marked for a decision.
Stage contract (Markdown)
The program charter metered
Pricing as soon as the session is ready...
Sign inThe reply could not be read. Here it is as it came back:
Caveats
Caveats
The page's check of this reply
Saved charters
How it works - and what is sourced
The shape follows the skill: an additive field on an existing pattern skips the beta; private preview is the default; design partners lead when the shape is still open or the feature brings a new resource model, the hybrid when GA is closer than twelve weeks; fewer than three qualified accounts fall back to design partners; no owner rules every shape out. Every stage carries eleven rows - an empty one is a promise you will be held to anyway. Personal data by nature needs test mode or synthetic data, not a ban; pre-GA terms override contracts promising "all services"; public preview carries an SLA only on a frozen shape; internal SLOs never sit in the terms.
The exit is written before the first account enrolls: a dated review at most twelve months out and before the GA target, a contract freeze at least one partner release cycle earlier, an activation floor, one extension at most. Each account needs evidence of the problem, a real integration in the window, a named technical owner, signed terms and - when billable - intent to pay; the cohort is capped by partner hours, and no partner may carry more than a third of the traffic. A beta that never ended is killed when usage is near zero, made GA as-is when its shape has been stable six months with production callers and no must-change finding, and otherwise moved to a next version with migration.