Settings and organization
Plan, usage and locks
What a lock means, the fifteen feature keys, the plan limits, usage per Environment, upgrade requests, and why locked docs pages show only a summary.
A plan decides which features an organization has and sets its limits. Reaching a limit does nothing to the data you already have.
What a lock means
Every plan-gated item stays in the left nav on every plan. When the plan lacks its feature, the item has a lock, and hovering names the plans that include it.
A locked item, or a link to a gated page, shows an upgrade panel in place of the page. It says what the feature does and has Request upgrade, See pricing and Back to where I was, which returns to the previous page. Locked buttons, such as Optimize on a trace, are disabled, with the reason on hover.
Features come from the best plan among all your organizations, and the locks and the plan in the sidebar footer both use it. The Plan pill on settings tabs and the Plan tile on project General show this organization's own plan, which can be lower.
The fifteen feature keys
| Key | What it covers |
|---|---|
datasets | Datasets: CSV and JSONL import, rows from traces, snapshots, synthesis, auto-add rules |
evaluators | Evaluators (library, custom judges, code, assertions, collections), runs, playground grids, prompts |
trace-evaluation | Evaluate on traces, online rules, calibration |
trace-optimization | Optimize on a trace, topics and patterns |
simulations | Agents, scenario sets, simulation runs, scoring, red team |
agent-publishing | Publishing an agent as a chat page anyone with the link can use |
cost-calculator | Estimate a workload |
project-text | Text generation surfaces: generate outputs, the Playground |
project-image | The Image tab of the Playground |
project-video | The Video tab of the Playground |
voice-eval | Voice datasets, the Voice tab, voice evaluation jobs |
chat-playground | The Playground chat, compare and prompts |
rl-labeling | Annotate: score types, review queues, labels, RL datasets |
alerts | Alerts and monitors |
benchmark | The public model benchmark board |
evaluators and project-text need datasets. trace-evaluation, voice-eval and simulations need evaluators, and agent-publishing needs simulations. The platform enforces these dependencies.
Traces, Cost, Settings, the notification bell, comments and end-user feedback need no key. Docs pages about a gated feature carry the same lock in the docs sidebar and show only their summary until the plan includes the feature.
Limits and usage
Plan & usage is under Settings → Organization in the sidebar footer; on the free plan, Upgrade beside the plan in the footer opens it too. It shows:
- the limits: Environments (
traceProjectLimit,-1for unlimited), trace retention in days (retentionDays) and the monthly span soft cap (spansPerMonthSoftCap) - spans this month for each Environment, summed in the browser over the Environments that answered
- the feature matrix, with the same lock icons as the nav
The soft cap is tracked, never enforced. Ingestion continues past it, and the organization Owner is emailed once at 80% and once at 100%, per Environment per month. At the Environment limit, Create environment opens the same upgrade panel, with the count, such as "1 of 1".
Request an upgrade
Any member can ask for an upgrade. On Plan & usage, Request upgrade asks for the Plan wanted (Pro or Enterprise) and an optional Message; on an upgrade panel, it requests the next plan at once. The platform team is emailed and replies within one business day.
Owners and Admins see the requests on Plan & usage as pending or closed. There is no self-serve billing. A platform admin changes the plan, the organization Owner gets a plan changed notification, and the nav updates on your next page load, since the plan cache refreshes within a minute.
A platform admin can also set a retention hold so the traces of an organization stop expiring.