Terms of use and disclaimer

inference.md is provided “as is” and “as available”, without warranties of any kind, express or implied, including warranties of accuracy, completeness, timeliness, availability, fitness for a particular purpose and non-infringement. You use it at your own risk.

inference.md shows no third-party data: no prices, rates, fees, market data or provider data, and it fetches nothing from other sites. Every number a tool uses comes from your request or from a generic formula the tool states. Those formulas are simplifications. They can be wrong for your case, and the skill and this site can be inaccurate, incomplete or not current.

Check every figure yourself, with the people and sources you rely on, before you act on it.

Nothing here is engineering, capacity-planning, financial, legal or other professional advice. Nothing on inference.md is a recommendation or endorsement of any provider, product or service. The INFERENCE.md skill is guidance for software agents and doesn’t guarantee any outcome.

The tools calculate, compare and check what you send against rules you choose. They don’t approve, enforce, certify or verify anything, including anyone’s identity, they can miss problems, and a clean result isn’t an approval or an assurance. Nothing on inference.md makes anything compliant with a law, regulation or standard, legally valid, or safe.

A pass from any tool means the numbers you sent met the targets you chose. It is not a service level agreement, a benchmark result, a certification or a promise about how any endpoint will behave. Results from a short test depend on prompts, lengths, load, network and time, and may not repeat. Availability can't be shown by a short test.

Capacity, replica and cost figures are estimates from your own inputs and from simple formulas such as Little's law. They are not quotes or prices, and real systems need more headroom than a formula shows.

Only test endpoints you own or have written permission to test, and stay inside their terms, quotas and rate limits. Tests cost money, can trip abuse controls and can affect other users. The local probe runs on your machine under your control; you are responsible for every request it or your agent sends, and for keeping your keys safe.

inference.md never calls your endpoints, runs tests for you or holds your keys. Anything your agents do runs on your own accounts, under the terms of the services you use, at your cost. You are responsible for the settings you choose, for approvals, for your credentials and for following every service’s terms and the law. Settings an agent drafted are yours only once you agree to them.

To the fullest extent permitted by law, the operators and contributors of inference.md are not liable for any direct, indirect, incidental, special, consequential or punitive damages, or for any loss of money, data, profits or business, arising from your use of, or inability to use, inference.md or any information from it, even if told such loss was possible. Where liability can’t be excluded, our total liability is limited to US$100.

The skill and the tools use generic terms and public, generic formulas. Names of standards, units and file formats are used only to describe them.

Where code on inference.md (for example, an install command or a configuration file name) mentions a tool, client or service so that it works with it, that name is a trademark of its owner, used only for compatibility. inference.md is not affiliated with, endorsed by or sponsored by any of them. If you own a mark and want us to change how we use it, write to us and we’ll act promptly.

Don’t overload, attack or try to disrupt inference.md, and don’t use it to break the law or anyone else’s terms. We may rate-limit, block, change or stop any part of inference.md at any time, without notice.

inference.md has no accounts, sets no cookies and asks for no personal information. To learn what’s useful, we count page views, copies, skill fetches and API and MCP calls with anonymous server-side events. We store no IP addresses: a one-way daily hash of the IP address and browser string counts unique visitors, and the country comes from our host’s request headers. Counting runs only on the live site, and visiting with ?internal=1 turns it off in that browser.

Settings and anything you send to a tool are used only to answer that request. inference.md doesn’t store them and doesn’t send their values to analytics: only the tool name, a known preset name and how many settings were given. Our host keeps standard server logs.

The INFERENCE.md skill and this site’s code are under the MIT License. Anything owned by third parties is not covered by that license.

We may update these terms. The date at the top shows the current version, and continued use means you accept it.