How agentplain may, and may not, be used.

Use agentplain to run your business. Don’t use it to attack the service, copy how it works, harvest what’s in it, or resell it. The rules below spell that out. We enforce them with automated detection and, where needed, account suspension — always with a notice to you and a way to appeal first.

Every account operates within per-account rate limits on requests, chat volume, connector reads, and model usage. The limits are sized generously for real business use and exist to keep one account from degrading the service for everyone else.

If you have a legitimate high-volume need, tell us — we’ll size a plan for it. What’s not allowed is engineering traffic specifically to find or exceed those limits (load generation, burst flooding, automated hammering of a single endpoint).

You agree not to attempt to extract, reconstruct, or reverse-engineer agentplain’s system prompts, skill definitions, vertical playbooks, insight library, or the internal orchestration logic that drives Plaino. This includes prompts engineered to make Plaino reveal its instructions verbatim, “ignore previous instructions” and jailbreak attempts, and systematic probing designed to map how the fleet decides and routes work.

Plaino is built to do your work, not to disclose how it is built. Our configuration — the curated vertical knowledge, the playbook logic, the persona — is our intellectual property and the result of substantial investment.

You agree not to use automated means (headless browsers, crawlers, scripts, scraping services) to harvest agentplain’s marketplace, knowledge corpora, product copy, or any other content from the marketing site or the product. Normal use of the product through its interface is fine; bulk automated extraction of what we’ve curated is not.

You agree not to run high-volume identity tests, credential-stuffing, or systematic probing against sign-in, sign-up, or any authentication surface; and not to flood the service with high-frequency identical queries intended to fingerprint, benchmark, or stress the underlying models rather than to do your own work.

You also agree not to abuse free trials — creating accounts to consume trial capacity, cancelling, and repeating under new identities is a violation, whether done manually or with automation.

Your subscription is for your own business’s use. You agree not to resell, sublicense, or repackage agentplain capacity to third parties; not to operate it as a service bureau on others’ behalf; and not to recreate agentplain’s orchestration model or vertical knowledge curation inside a competing product.

You agree not to decompile, disassemble, or otherwise attempt to derive the source code, model weights, or internal architecture of agentplain or Plaino, except to the limited extent that applicable law expressly permits despite this restriction.

You also agree not to use agentplain to break the law in your vertical; to send communications that violate CAN-SPAM, the TCPA, GDPR, or similar laws; to misrepresent yourself to recipients of drafts; to attempt to access another customer’s data; or to probe, scan, or test the vulnerability of our systems without our written authorization.

We monitor access patterns for abuse — and for nothing else. (We do not use your activity to train models; see our privacy policy.) When our systems detect a likely violation, we may soft-suspend the account: the workspace drops to read-only, we email the account owner with the reason, and a review window opens during which you can appeal. Your data stays intact and exportable the entire time.

If a violation is confirmed, we may fully suspend the account. Even then your data is preserved per the retention terms in our terms of service, and you can still request an export. We aim to be proportionate: a first, accidental trip of a rule gets a notice, not a ban.

Material changes are announced by email to your workspace owner at least 30 days before they take effect. The latest version always lives at this URL.

Contact: hello@agentplain.com. See also our terms, privacy, and security pages.