The New Limitr Cloud: Limitr Analyst, Custom Contracts, and Pay As You Go

The new Limitr Cloud is live at cloud.limitr.dev. It's a rebuilt app and a rebuilt server, running the 0.7 engine, with Limitr Analyst, the agent in Limitr Cloud, working beside you on every page. It also comes with new plans, including Pay as you go.

When we shipped Limitr 0.7, the engine could price every call at its real cost, predict a call's size before it runs, roll features out, and carry custom deals per customer. Cloud is where your team sees all of that and changes it: one policy, every customer, and what each of them costs you and earns you.

One policy, managed in one place

Everything about how you price and control usage lives in one versioned policy: plans, entitlements, credits and their prices, limits, credit packs, and the rules for custom contracts. Cloud publishes it to the Limitr engine in your app, which decides every call in your process, with no network call on the decision path.

• A shared draft. You, your teammates, and the Analyst edit the same draft. Nothing reaches your app until you publish.
• Tables or code. Edit entitlements, plans, credits, and pricing rules in tables, or the whole policy in the Code tab, checked as you type with each problem marked on its line.
• Review, then publish. The Changes tab shows exactly what moved since the last publish. Publish to test first, copy test to live, and roll back to any earlier version.
• Notes that travel with the policy. What you sell, your goals, and the decisions behind your pricing live in the policy's notes, versioned with it and never sent to your app. The Analyst reads them before it changes anything.
• Resets that match the bill. A limit or credit pack can start over with each customer's billing period, as their invoice closes, instead of on the 1st for everyone.

Limitr Analyst

Limitr Analyst works with you inside the app. Ask it in plain words ("add a 14-day trial to Pro", "which customers are over their limit this month", "show me margin by model for the last two weeks"), and it already knows which page, workspace, and mode you're on.

• It writes your policy in the shared draft and says what it changed, then checks its work on made-up customers before it tells you what the policy does.
• It reads your data: customers, usage, events, and segments.
• It shows you. When a picture answers better, it draws a view in the conversation: usage over time, cost by model, margin per task. Open it full width, or save it as a dashboard for your team.
• It proposes, you apply. Outside the draft, the Analyst never changes anything on its own. Moving a customer to a new plan, giving them more of a feature until December, a new segment, or a publish arrives as a card with each change's before and after, and nothing happens until you press Apply. It applies as you, with your permissions.
• It shares. Share a conversation with your team, and a teammate can read it or carry on from it in their own conversation.

Two of the Analyst's jobs are new, and they're where pricing work actually gets hard. Both are on Enterprise, in early access:

• Forecasts and what-ifs. Usage, revenue, and margin ahead, who will hit a limit this period and when, and a draft's prices run against your customers' real bills before you publish it, so you see what a change does to each customer and to margin.
• Contract creation. Hand it a deal's terms and it drafts them into your policy: the customer's plan, their own limits and prices, and the pricing rules around them, ready for you to review.

Customers, margin, and billing

• Every customer, and what they're doing. Their plan, usage, limits, credit balances, caps, and this period's bill, with every change you can make to one customer: a plan, a limit of their own, giving or taking a feature, a cap, a credit pack, their own prices, and how they're billed.
• Margin as you go. Dashboards show usage, cost, value, and margin from your calls, broken down by the metadata your calls carry. Group calls by a key like task_id and you see what a whole task cost and what it earned, across every model and tool it used. Every number says how it was calculated.
• Segments and notifications. Group customers by plan, metadata, activity, or usage, filter any page by a segment, and send alerts and email when something happens: a limit reached, overage, a cap, a plan change.
• Invoicing on your Stripe. Cloud keeps each customer's usage ledger and closes their invoices every period. Connect your own Stripe to charge them, or mark a plan or a customer as billed manually when they pay by ACH, wire, or PO.
• Your team. Invite teammates and choose what each can do: publish, change customers, manage keys, the team, or billing.

Connecting your app

Your app runs the engine through the SDK, and Cloud keeps its policy and customers in step:

import { Limitr } from '@formata/limitr';

const limitr = await Limitr.cloud({ token: process.env.LIMITR_KEY! });

await limitr.ensureCustomer('org_acme', 'pro');
if (await limitr.allow('org_acme', 'chat', tokens, { model, task_id })) {
    // allowed, metered, and in your dashboards
}

The SDK keeps the engine it ships with, and Cloud sends it your policy and customers, so a publish reaches every running copy of your app on its next call, with no release. Usage goes back to Cloud in batches. If the connection drops, your app keeps deciding with what it has.

The easiest way to integrate is your coding agent. With the Limitr skills (in Claude Code, /plugin marketplace add dev-formata-io/limitr, then /plugin install limitr@limitr), ask it to connect your app, gate an endpoint, or build a usage meter for your account page.

Plans

Limitr Cloud's own plans run on Limitr: every workspace is a customer of ours, metered by the same engine you use.

• Free. The policy editor, starter templates, drafts and validation, and Limitr Analyst with $1 of usage to try it: it writes your policy and tests it on made-up customers. No card needed.
• Pay as you go. $49 a month for teams going live on their own: publishing and version history, API keys and connected apps, customers, segments, notifications, invoicing with Stripe, dashboards, and the Analyst with your data. It includes 50 active customers (a live customer with a governed call that month), then $0.50 each, and 2,000 calls per active customer, then $0.02 per 1,000. The Analyst's first $5 of model usage each month is included, then it's billed at cost plus 20%. A monthly budget alert emails you before a bill surprises you, and it never pauses anything. Test mode is never billed (it's for building and testing, up to 250 test customers and 250,000 test calls a month), and it's never a share of your revenue.
• Enterprise. An annual contract for teams selling custom contracts or running at volume: everything in Pay as you go, plus pricing rules and per-customer prices, contract creation and forecasting with the Analyst, the Analyst on the largest model for deep work, saved dashboards, usage included by contract, and direct support from our team on pricing, usage control, and margin. Past a few thousand active customers, it costs less than Pay as you go.

Details and an estimator are on the pricing page.

Engine 0.7.5

The engine moves to 0.7.5 with this release. It adds reset_sch: 'period', so limits, credit packs, and caps start over with each customer's billing period, and events when an exchange rate changes at runtime, which Cloud records. The 0.7 post has the details.

If you used Limitr Cloud before

Your workspaces, policies, customers, and teammates came over, and your app kept running through the move. Apps on SDK 0.6 keep working for now, but update to @formata/limitr 0.7.5 to get everything above. Support for 0.6 ends in November.

Get started

Start free and have the Analyst write your first policy, or talk to sales about custom contracts, volume, or an Enterprise trial. The engine is open source on GitHub, and the Cloud guides walk through every page of the app.