Terms of Service

Template / draft. Published for review; not legal advice and not yet counsel-reviewed. Have a qualified attorney finalize before you rely on it.

1. The service

Outlay ("we") provides a hosted console and read-only integrations that attribute your LLM and coding-agent spend to the work that drove it, forecast it by scope, and help you hold it to budget ("Service"). The Service connects read-only to your work tracker and your model provider's usage/billing data. Optionally, you may run our opt-in enforcement gateway in front of your own model calls to enforce a budget you set (blocking or routing down over-budget traffic); that gateway is the only component that sits in your request path, it is entirely opt-in, and it fails open. You ("Customer") use the Service with your own model-provider account(s) and API key(s); we are not a party to your agreement with any model provider and do not resell model usage.

2. Accounts & acceptable use

You are responsible for your account, your users, and keeping credentials secure. Your use must comply with our Acceptable Use Policy, applicable law, and Anthropic's usage policies.

3. Fees

Fees are as set out in the order form, quote, or written agreement between you and us. Pricing is scoped to your usage and agreed before you are charged; any pilot or evaluation period is free unless otherwise agreed in writing. If no order form or written agreement is in place, the Service is provided for evaluation only and no fee is owed. Fees are billed via our payment processor, and taxes are your responsibility where applicable.

4. Estimates & forecasts disclaimer

The spend attributions, forecasts, and estimates the Service produces are estimates measured from your own data, not guarantees. Actual results depend on your workload, prompts, cache state, model pricing, and how your work is tracked. We do not warrant any particular figure, forecast accuracy, or savings amount; you remain responsible for evaluating the Service's outputs for your use case.

5. Availability

The Service's read-only attribution and forecasting do not sit in your request path — removing the Service does not change how your model calls are made. The optional enforcement gateway is designed to fail open: if it is unreachable, requests pass through to your model provider rather than being blocked. We do not guarantee uninterrupted operation and, except where a separate written SLA applies, the Service is provided without uptime commitments.

6. Warranties

THE SERVICE IS PROVIDED "AS IS" AND "AS AVAILABLE" WITHOUT WARRANTIES OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, AND NON-INFRINGEMENT, TO THE MAXIMUM EXTENT PERMITTED BY LAW.

7. Limitation of liability

TO THE MAXIMUM EXTENT PERMITTED BY LAW, WE WILL NOT BE LIABLE FOR INDIRECT, INCIDENTAL, SPECIAL, CONSEQUENTIAL, OR PUNITIVE DAMAGES, OR LOST PROFITS/DATA. OUR TOTAL LIABILITY FOR ANY CLAIM IS LIMITED TO THE FEES YOU PAID US IN THE THREE (3) MONTHS BEFORE THE EVENT GIVING RISE TO THE CLAIM. You are responsible for your own Anthropic/model-provider charges. (This three-month cap applies to self-serve customers on these Terms; customers under a negotiated MSA have a separate, twelve-month cap — the difference is intentional.)

8. Data

Our handling of data is described in the Privacy Policy and, for personal data processed on your behalf, the DPA. In the default read-only deployment, prompt content, model outputs, and your API keys are not transmitted to us — the Service reads usage metadata and work-tracker references only. If you opt into running the enforcement gateway, your API key and request content are processed in memory solely to enforce the budget you set and are not persisted — handled as your processor under the DPA; we do not store prompt content or model outputs.

9. Term & termination

You may stop using and cancel at any time. We may suspend or terminate for breach (including AUP violations) or non-payment. Sections that by their nature should survive (fees accrued, warranty disclaimers, liability limits) survive termination.

9a. Early access

The Service is offered as an early-access product. Features may change, and we may adjust or deprecate functionality as we improve it; we'll communicate material changes. Enterprise customers can request a negotiated MSA with support and (optionally) SLA terms.

10. Your Anthropic agreement & related disclosures

You use the Service with your own Anthropic account and API key, under your own agreement with Anthropic (their Commercial Terms, Usage Policy, and Service Specific Terms). We are not a party to that agreement and do not resell model access. You remain responsible for complying with it, including:

Usage Policy compliance flows through to you. Anthropic's Usage Policy applies to all inputs, including those reaching the model through a passthrough/proxy. Where you operate a High-Risk Use Case (e.g., legal, healthcare, insurance, finance, employment or housing, academic testing, or journalistic content), Anthropic requires a qualified human in the loop and an AI-involvement disclosure to your end users. Any consumer-facing chatbot or interactive agent must disclose to users that they are interacting with AI. You are responsible for meeting these requirements.
Anthropic's IP indemnity is governed solely by your agreement with Anthropic. Your agreement with Anthropic — not Outlay — determines the scope of any copyright or IP indemnity Anthropic provides, including any exclusions (for example, exclusions that can apply where outputs result from modifications to, or combination of, the Services with technology not provided by Anthropic). Outlay makes no representation or warranty as to whether any feature of the Service does or does not affect that indemnity. Because the Service offers optional features (e.g., model substitution, response caching, semantic caching, fallbacks) that you can enable or disable, you are responsible for determining — and confirming directly with Anthropic — whether your configuration affects your coverage. This paragraph is a disclosure for your awareness; it is not legal advice and is not a statement about your coverage.
Anthropic's own data handling is outside our control. Our privacy commitments describe only what Outlay does (in the default self-hosted deployment, the proxy does not transmit your prompts, outputs, or API key to us). Anthropic's handling of your data is governed by your Anthropic agreement — for example, Anthropic may retain inputs and outputs for safety review on certain models even where zero-data-retention would otherwise apply, and Anthropic's Development Partner / data opt-in setting permits training on your content if you enable it. Do not enable that setting unless you intend that result; we never enable it on your behalf.

11. Changes; governing law

We may update the Service and these terms; material changes will be communicated. These terms are governed by the laws of the State of Delaware, without regard to its conflict-of-laws rules, and the parties submit to the exclusive jurisdiction of the state and federal courts located in Delaware (final venue language to be confirmed in counsel review). If any provision is unenforceable, the rest remain in effect.

Questions: [email protected] · Privacy · Acceptable Use · DPA