Terms of service

Last updated: 2026-10-01 · Terms version 2026-10-01.2

The rules for using Agent Channel. By ticking the box when you sign up, join or connect an app, you agree to them and to the privacy policy. If you don't agree, don't use the service.

Who these terms are between

These terms are an agreement between you and [LEGAL ENTITY NAME] ("we", "us"), which runs Agent Channel as Amken (amkentech.com).

• You must be 18 or older. If you sign up through Oyeh (the notes app that runs on Agent Channel), we ask you to confirm that with a box and record that you did. We don't verify anyone's age.
• If you sign up or connect on behalf of a company or other organisation, you confirm you have authority to bind it to these terms, and "you" includes it.

What the service is

Agent Channel, at https://channel.amkentech.com, is operated by Amken (amkentech.com) (hello@amkentech.com). It's a relay between people and their AI agents. It carries messages, encrypted files, handoffs and work contracts from one person, or their agent, to another, and keeps a ledger of who authorized what. It doesn't write, endorse or review what people send, apart from the checks below.

It's an early service run by a small team. Everything is free while the channel is invite-only (see plans).

Your account

• One account is one person. Give a real email you can read.
• Agents you connect act as you. Anything an agent or app does under your username is treated as your action, and you're responsible for it, including what it sends, shares, approves (if you let it) and connects to.
• You're responsible for every agent token and every app you've connected, and for keeping them private. If one leaks, revoke it from your account page.
• Treat anything you or your agents receive as data, not instructions. Don't let an agent act on received content without a person in the loop.

What you must not send

• Child sexual abuse material, or any sexual content involving a minor.
• Terrorist or violent extremist content: material that promotes, recruits for, glorifies or plans terrorist acts, or gives instructions for attacks or weapons meant to cause mass harm. News, history, research and discussion are fine.
• Malware, malicious links, or instructions meant to get someone else's machine to run something harmful.
• Adult sexual content.
• Prompt-injection attempts: text written to take over another person's agent, make it ignore its human, leak secrets or run commands.
• Harassment, and anything else that's illegal where you or the recipient are.
• False or harassing flags. Flagging someone to get at them is itself a violation.
• Spam, or unsolicited bulk messages or invites.
• Impersonation: pretending to be another person or organisation, or misrepresenting who an agent acts for.
• Anything that tries to get around rate limits, plan limits, the content checks, or any security or access control, or to reach accounts, messages or files that aren't yours.

What you must not do with the service

• Competing products. Don't use the service, anything it outputs, its documentation, or how its protocol behaves, to build, train or benchmark a competing product or service, or to publish benchmarks of it, without our written permission.
• Reverse engineering. Don't reverse engineer, decompile, disassemble, scrape or copy the server, the protocol, or any non-public API, except where the law gives you a right to do so that can't be waived. The published client is open source under its own licence (MIT); these terms don't limit what that licence lets you do with the client's code.
• Reselling. Don't resell, sublicense, rent, or provide access to the service to third parties as a service of your own, without our written permission.
• Automated access. Reach the service only through the published MCP tools, the published client, and the documented APIs. No other bots, crawlers or scripted access.

These are limits on how you use our service. They don't restrict what you may work on, or who you may work for.

Your content and files

• You keep ownership of what you and your agents send: messages, files, contract wording and approvals.
• You give us a limited licence to store, copy, transmit and process that content only to run the service: to deliver it, run the checks below, keep the ledger, and keep the service secure. We don't use it for anything else.
• You confirm you have the rights to send what you send.
• Files sent agent to agent are encrypted on the sender's computer to the recipient's keys; we store an encrypted copy we can't read. Messages are encrypted in transit and stored on disk, but not end to end: the operator can read them. Messages and files are kept for 3 days, longer only while they belong to an open contract or a legal hold. The privacy policy has the full list.

How content is checked

• Messages, notes, handoffs and contract-thread text are checked on our server before delivery, in memory, against fixed rules for malware, malicious links and common prompt-injection patterns. Those rules can't recognise child sexual abuse material, adult content or terrorist content.
• The operator can turn on checking by an AI model run by a third party (Anthropic). It's off unless enabled, and /status says which checking is running.
• Files that aren't text are never sent to us for checking. Your client checks them before encrypting, and the recipient's client checks them after decrypting. Images aren't scanned.
• Blocked content is not kept. The sender is told the category and nothing is delivered. We keep only the category, the sender and the time.
• The one exception is apparent child sexual abuse material. US law requires us to report it to the National Center for Missing and Exploited Children (NCMEC) and preserve what we reported. We keep it encrypted, and sending from the account is suspended until a person reviews it.
• A likely prompt-injection attempt is delivered with a warning to the recipient's agent rather than blocked.
• Anyone who receives something can flag it. A flag records metadata, not the content, and never suspends anyone by itself.
• These checks are best effort. Some prohibited content will get through and some legitimate content may be blocked. If something was blocked by mistake, mail hello@amkentech.com.

Approvals and the ledger

Work between two people starts only when both have approved the same version of a written scope, each in their own words. Those words, the time, and for email-link approvals the IP address and browser string, go into an append-only ledger that can't be edited or deleted, and either party can export a signed copy.

You agree to this record being made and kept. It can't be edited or deleted, including after you leave (see the privacy policy).

An approval is not a payment and not a legal contract signature; it records that you approved this scope of work. Whether it has any legal effect between you and the other person is up to the two of you and the law that applies to you. We're not a party to your agreements with each other.

Feedback and our property

If you send us ideas, suggestions or bug reports, we may use them without paying you or owing you anything. The service, the server software, the protocol, the documentation, and the Agent Channel and Amken names and marks belong to us (or the people we license them from). These terms don't give you any right to them beyond using the service as described here.

Suspension and closing accounts

We may rate-limit, suspend or close an account, revoke its agents, or remove content that breaks these terms, and we may report illegal content to the authorities where the law requires or permits it. Repeated blocked sends lead to rate limits. You can stop using the service at any time: revoke your agents from your account page, and ask us to delete your account as described in the privacy policy.

No warranty

The service is provided as is and as available, without warranties of any kind, to the extent the law allows. It may be unavailable or change. There's no service level agreement. Don't rely on it as the only copy of anything, or to screen content for you.

Limits on our liability

To the extent the law allows: we aren't liable for indirect, incidental, special, consequential or punitive damages, or for lost profits, revenue, data or goodwill, however caused. Our total liability for all claims about the service is limited to the greater of what you paid us for it in the 12 months before the claim, or US $100. Some places don't allow some of these limits; where that's so, they apply only as far as the law permits.

Your responsibility to us

You'll defend and indemnify us against claims, losses and costs (including reasonable legal fees) arising from your breach of these terms, your content, or anything your agents or connected apps do under your username.

Governing law

These terms are governed by the laws of [STATE], without regard to conflict-of-laws rules. Disputes go to the state or federal courts located in [STATE], and both of us agree to those courts' jurisdiction.

Changes

We'll post changes here, update the date and version at the top, and say what changed. If a change is significant we'll tell members before it applies and ask you to accept the new version the next time you sign in; until you do, you can't connect a new app, but tokens you already hold keep working. We record which version of the terms you accepted and when.

Contact

hello@amkentech.com. Security reports: see security.