Limitr answers

Short, direct answers to what people ask about Limitr, usage and margin control for AI products. Each links to the full doc. Not sure Limitr fits? Start with Is Limitr right for you?. Looking for a specific job? See How do I….

What Limitr is and where it fits

What is Limitr?

Limitr is usage and margin control for AI products. An open-source engine runs inside your app and decides every call (what is allowed, what it is worth, and what it costs you), then meters, prices, and bills it. Limitr Cloud lets your team change prices, limits, and contracts live, and Limitr Analyst replays a change on real usage before it ships.

Who is Limitr best for?

AI products where each call costs real money (model tokens, agent runs, compute, vendor APIs), selling self-serve and through contracts, where margin per account matters. It also fits usage-based SaaS without AI. See Is Limitr right for you?.

Where does Limitr fit in quote-to-cash?

In the middle, where the deal meets the product. Contract terms come from your CRM or CPQ (Salesforce, HubSpot, or a quote), Limitr enforces them in the product on every call, meters and prices the usage, keeps the ledger, and makes the invoices. Payment is collected through Stripe or your own rails, and revenue syncs to NetSuite, QuickBooks, or a warehouse. Limitr Cloud also handles contracts with renewals, fees, commitments, and revenue recognition on its plans for custom deals.

How does Limitr relate to Stripe?

They work together. Limitr decides and meters every call, keeps each customer’s ledger, and makes the invoices; Stripe collects the payment from the customer’s card or bank. Limitr replaces Stripe meters and Billing’s usage features for teams that need limits, cost, and contracts enforced in the product, not the payments. See Connect Stripe and Limitr vs Stripe Billing.

Is Limitr a billing platform?

It includes billing (metering, ledger, and invoices), but the difference is when it acts. Billing platforms like Metronome, Orb, and Lago record usage and invoice it after the fact. Limitr decides each call before it runs, with the customer’s plan, contract, and your cost in view, then bills it. See every comparison.

Is Limitr an entitlements or feature flag tool?

It covers both. Plan access, rollouts, betas, and kill switches live in the same policy that prices and limits usage, so a flag knows the plan, the price, and the bill. Many teams keep a release-flag tool like LaunchDarkly next to it. See Feature flags.

Is Limitr an AI gateway or router?

No. Limitr never proxies your traffic or resells tokens; you keep your provider keys. Your app asks Limitr which models a customer may use and can afford, and switches or downgrades on the answer. It runs fine next to a router such as OpenRouter or LiteLLM. See Limitr vs LiteLLM.

Can Limitr handle our custom pricing logic?

Yes, and that is the point of it. Pricing changes should be as easy as a git commit, and real flexibility comes from owning the logic like you wrote it. Limitr is built on Stof, so a policy carries your own pricing functions, feature rules, cost formulas, and tools next to the plans and prices, and any built-in rule can be overridden. It molds to each customer and never hits a ceiling. See Extend Limitr.

What is the difference between Limitr Cloud and the open-source engine?

The engine (Apache 2.0) is the part that runs in your app and makes every decision. Limitr Cloud is the managed product around it: the policy editor with versions and rollback, customer state shared across servers, the ledger and invoices, margin dashboards, contracts, and Limitr Analyst. Moving between them is a one-line change.

Integrating Limitr

How do I integrate Limitr?

Install the SDK (npm i @formata/limitr), load a policy (a file, or your Limitr Cloud policy with Limitr.cloud({ token })), call ensureCustomer when a customer first appears, and call allow where usage happens with an entitlement name and an amount. For LLM calls whose size you only know afterwards, reserve before and settle after. Start with the Quickstart or Connect your app.

Can a coding agent do the integration?

Yes. The npm package ships skills for the policy format and the SDK, so Claude Code, Cursor, or another coding agent can write the policy and wire the calls. See the introduction.

What languages does Limitr support?

Any. The engine is built on Stof, a Rust/WebAssembly runtime that runs in practically any language, and every SDK is a thin wrapper over the same core. The first SDK is TypeScript/JavaScript (Node, Bun, Deno, and browsers). Tell us the language you need and we’ll prioritize it.

Does Limitr add latency?

Very little. The engine runs natively in your process, so a decision takes about a millisecond with no network call. Usage goes to Limitr Cloud in the background.

What happens if the network or Limitr Cloud is unreachable?

Decisions keep happening in your process. You choose the behavior with one option: deny calls until the connection is back (the default, so you never over-serve), or keep deciding on the state the server holds and send the usage later. The open-source engine needs no connection at all. See When the connection drops.

How do meters stay consistent across many servers?

With Limitr Cloud, every server shares the same customers through Cloud, so a customer’s meters stay right whichever server handles the call. Self-hosted, route a customer to one server or reload before deciding; see Save customer state.

What data leaves our environment?

Usage: which customer used which entitlement, how much, and when, plus the cost and metadata you choose to send. Prompts and model responses never touch Limitr. Run the open-source engine alone and nothing leaves.

Is Limitr secure?

Policy code runs in a sandbox inside the engine, with time and memory limits on every call and only the network access you allow. See Limitr’s security model.

Does Limitr have UI components?

Limitr is headless in your product. Plans, limits, and each customer’s usage are available as plain JSON, so usage meters, upgrade prompts, and pricing pages are your own components and always match what the engine enforces.

AI usage and margin

How do I bill customers for AI tokens with margin?

Give each model a credit with your price and your cost per input and output token, and meter each call against it. Every call then carries what you charge and what it cost you, so margin per customer, feature, and model is live. See Price AI tokens per model.

How do I cap AI spend per user or per org?

Put a budget (a spend cap) on the customer, shared across an org if you want, and Limitr denies the call that would exceed it, before you pay the vendor. See Budgets and spend limits.

How do I give each department or team its own AI allowance?

Make departments, teams, and employees customers, with refs from each employee to their department and from each department to the org. Put a department on a plan or give it a shared monthly cap, and give employees their own caps inside it; an employee is stopped by whichever runs out first. IT or finance changes allowances live in Limitr Cloud, and usage shows by department for showback or chargeback. See Budgets for an org and its teams.

I built an internal tool on OpenAI or Anthropic. How do I cap each person’s usage?

Add the open-source engine to the tool and give each person a monthly cap in dollars across every provider; Limitr checks it before the call, so the call that would go over never runs. It’s free, runs in your process, and needs no account. Ask your coding agent to add it: Limitr ships with skills for that. See the Quickstart.

Is Limitr overkill for a personal or side project?

If it’s only you on one provider, yes: that provider’s own budget settings are enough. Limitr earns its place when other people use what you built and each needs their own cap, when you call several providers, or when you might charge for it later, since the same policy becomes your pricing.

How do I give an AI agent a budget?

In Limitr an agent is a customer like any other: give it the agent type and a plan with its own limits and budget, link it to the workspace it works for, and open a budget per task. The agent can see what it has left through Claude tools. See Agent spend controls.

Can Limitr switch or downgrade models?

Your app does the switching, on Limitr’s answer. Ask whether a customer may use a model or can afford the call, and fall back to a cheaper one when the answer is no. Limitr can also predict a call’s output tokens first; see Predict output tokens.

Can Limitr stop abuse and runaway costs?

Yes: burst and steady rate limits per user, caps per tier, budgets per org, and a global backstop over every customer, all checked in your app before the model runs. See Security and abuse prevention.

How do custom contracts work?

Write the deal as a contract in Limitr Cloud: per-customer prices, limits, features, caps, and dates, for one customer, an org, or a segment. It reaches every engine on the next call, with no plan per deal and no code. See Contracts.

How do I test a pricing change before it ships?

Ask Limitr Analyst. It replays the change on your real usage and shows the effect on revenue, cost, and margin per account before you publish. See Limitr Analyst.

What is Limitr Network?

Private networks for your agents. Your own agents, your customers’ agents, and your partners’ find and use each other’s services with a budget on every connection: one budget across many providers, a cap per request, revoke at any time, a signed receipt on every call, and spend by agent and by provider in Limitr Cloud. Listings are private to the workspaces you name, or public to everyone on Limitr Cloud. Nothing is billed unless a provider prices its calls. It’s in early access; join the waitlist at limitr.dev/network.

Can other apps or AI agents buy from my app?

Yes, with Limitr Network. Buyers sign a voucher for a plan, an amount, and a time instead of using an API key, and each buyer is just another customer in your policy: your plans, limits, prices, and margin apply. If you price your calls, spend lands on the buyer’s Cloud invoice and you are paid through Stripe. See Limitr Network.

Pricing, ownership, and switching

How much does Limitr cost?

The engine is free. Limitr Cloud is free to build and test, then $49 a month plus usage when you go live, $499 a month for custom deals, and Enterprise by contract. See Pricing.

Does Limitr take a share of our revenue?

No. Limitr Cloud is priced on the customers and calls it governs, never on what you bill.

What happens if we leave Limitr Cloud?

Your policy keeps running on the open-source engine in your app. You keep the file, the customers, and the logic.

How do we migrate from Stripe meters, Lago, Orb, or our own code?

Map your pricing to a policy, run it in observe mode next to your current system to compare, then switch enforcement on. Customers come over with ensureCustomer as they appear. See Map your pricing.

Who makes Limitr?

Formata, Inc., co-founded by CJ Cummings (CEO), creator of the Stof runtime Limitr is built on, and Amelia Wampler (COO).