Acceptable Use Policy

1. General use

You agree NOT to use Reach:

• To violate a third-party site's terms of service. Reach surfaces capabilities declared in /.well-known/ manifests; if a target site's terms of service forbid the action you intend to invoke, do not invoke it through Reach.
• To exceed rate limits. Reach honors rate limits declared in manifests. Do not use Reach to launch high-volume or denial-of-service traffic patterns against any site.
• To invoke capabilities you are not authorized to invoke. Deleting another person's data, transferring money out of an account that is not yours, or otherwise taking actions on someone else's behalf without their authorization is misuse of Reach.
• To perform automated scraping beyond the rate and scope declared in a target site's manifest. The shadow registry's purpose is to surface declared agent capabilities, not to provide a bypass for sites that have not authorized automation.
• To circumvent paywalls, ticketing-system fairness controls, or anti-bot measures that the target site uses to limit automation.

2. Shadow registry contributions

Anyone may contribute to the shadow registry by opening a Pull Request (PR) at github.com/inferlane/reach. Contributions must meet the following standards:

2.1 Accurate sourcing

Every contributed manifest must cite the public Application Programming Interface (API) documentation it adapts. Manifests reverse-engineered from undocumented internal endpoints are NOT accepted. The shadow registry exists to make declared interfaces more accessible — not to expose interfaces that the target site has not chosen to publish.

2.2 Correct sensitivity classification

Every capability must be tagged with one of three sensitivity flags:

• read: idempotent retrieval — for example, GET /repos/{owner}/{repo}/issues.
• write: creates or modifies state — for example, POST /repos/{owner}/{repo}/issues.
• destructive: permanent state change, payment movement, account deletion, refund issuance, subscription cancellation.

When in doubt, classify higher. Misclassification of a destructive action as "read" can cause real harm to users who trust the sensitivity flag. Reviewers will reject PRs that downgrade obvious destructive actions.

2.3 Honoring domain opt-outs

Domain owners may opt out of the shadow registry by publishing a file at /.well-known/reach-optout on their domain. The file is a plain-text file containing the single word OPTOUT. The presence of this file means:

• Reach will NOT return shadow-registry entries for that domain at runtime.
• Existing shadow-registry entries for the domain will be removed via a follow-up PR.
• New PRs for opted-out domains will be rejected without further review.

If you operate a domain and want to opt out without waiting for shadow registry maintainers to discover the file, email registry@inferlane.dev with the subject "Opt-out: {your domain}." We will remove the entry within seven (7) days.

2.4 Intellectual property

By contributing to the shadow registry, you represent that you have the right to do so. You grant InferLane and other users a perpetual, irrevocable, royalty-free license to use, modify, and distribute your contribution under the MIT license. Do not submit manifests that copy text from a target site's proprietary documentation verbatim; paraphrase descriptions in your own words.

3. Prohibited content

The following content is prohibited in the shadow registry, in InferLane support correspondence, and in any communication with InferLane:

• Content that promotes or facilitates illegal activity (including but not limited to malware, fraud, child sexual abuse material, terrorism, doxxing, or stalking).
• Content that infringes a third party's intellectual property rights or trade secrets.
• Content that misclassifies sensitivity flags to mislead end users.
• Content targeting financial accounts, medical records, or other regulated data classes where the contributor cannot show authorization.
• Spam, advertising, or affiliate-link injection.

4. Premium SLA adapters

Premium adapters are maintained by InferLane staff and are not subject to community contribution. The following additional rules apply:

• InferLane warrants the adapter's uptime per the order form's SLA. Refund triggers and credits are defined there.
• Reselling premium adapter access is prohibited. Each subscription is for a single organization's use.
• Reverse-engineering, scraping, or otherwise extracting premium adapter content for republication is prohibited.

5. Reporting violations

To report a violation of this AUP, email abuse@inferlane.dev. Include enough detail to investigate: the domain involved, the registry entry, a screenshot if applicable, and your relationship to the matter (domain owner, end user, third party).

InferLane will acknowledge the report within five business days and act on substantiated violations promptly. Repeat violators may be permanently barred from the hosted services.

6. Enforcement

InferLane may, at its discretion and without prior notice in cases of immediate harm:

• Suspend or terminate access to the hosted ChatGPT endpoint for the account responsible.
• Remove shadow registry entries that violate this AUP.
• Permanently bar repeat violators from contributing.
• Cooperate with law enforcement when legally required.

7. Changes

InferLane may update this AUP. Updates take effect on the date posted. Material changes will be announced via the inferlane.dev landing page banner.