Disclaimer

The tools on this site are useful and the guides are carefully researched. Neither is authoritative, and neither is a substitute for checking the primary source. Here is precisely where each one stops being reliable.

Token counts are estimates

The AI token counter uses a BPE-style approximation. On ordinary English prose it is typically within about 10% of the exact count. On code, JSON, numbers, non-Latin scripts and emoji it is less reliable.

Do not use it for billing, invoicing, or capacity guarantees. For an authoritative count, use tiktoken, your provider's count_tokens endpoint, or the usage object returned on every API response. That last source is what you are actually charged against.

Pricing data goes stale

Any price shown on this site was correct when checked and carries the date it was verified. Model lineups and prices change frequently — new generations ship every few months, and rates move with them.

Always confirm on the provider's official pricing page before committing to a budget, a contract, or an architecture. Where sources disagreed, we omitted the figure rather than publish a number we could not stand behind.

Security guidance is general

The JWT decoder tells you what a token contains and flags common problems — expired tokens, alg: none, missing exp. It does not and cannot verify signatures, since that requires your key.

Security guidance here is general best practice, not an assessment of your system. It does not replace a threat model, a code review, or a penetration test by someone with access to your actual infrastructure.

Not professional advice

Nothing on this site is legal, financial, security, or medical advice. It is general technical information for developers. Decisions with legal or financial consequences should be made with qualified professionals who have access to your specific circumstances.

Tool limitations

• YAML conversion supports the common structural subset. Anchors, aliases, merge keys, multi-document files and custom tags are not fully supported.
• SQL formatting tokenises rather than fully parsing. Deeply nested CTEs, window functions and vendor-specific syntax may indent unexpectedly.
• The regex tester uses JavaScript regex semantics. Other engines differ in details — test in your target language.
• Cron next-run previews assume standard five-field cron semantics. Your scheduler may differ; some platforms add seconds or year fields.
• The diff tool diffs by line and uses an algorithm whose cost grows with the product of the two input sizes, so very large inputs are refused.

Third-party advertising

Advertisements are served by Google and its partners. Their presence is not an endorsement by us, and we do not control or vet their content. Advertiser claims are the advertiser's own.

External links

Links to documentation and vendor pages are provided for convenience. We do not control those sites, are not responsible for their content, and cannot guarantee they remain accurate or available.

Corrections

Accuracy matters to us. If you find something wrong — a price, a factual claim, a tool that mishandles your input — please tell us. Corrections are prioritised and made quickly.