From: Kushal <kushal@kushalsm.com>
Date: Mon, 18 May 2026 05:03:11 +0000
Saw your question on the Agent Vault thread about websocket-frame auth
(Home Assistant) and the worry about the model reflecting the bearer
token back into its own context.
chrome-relay's answer is structurally different: the credential never
enters the agent's context because the agent never touches it — the HA
session lives in your real Chrome (cookies, WS handshake and all), and
the agent drives the tab over CDP, only ever seeing the rendered page.
URL: https://chrome-relay.kushalsm.com/
For your HA + agent setup today, are you keeping the session alive in a
browser the agent attaches to, or doing the WS auth on the agent side
and managing the token-in-context risk yourself?
Kushal
Read to me like an LLM had written it. It references something I said in a HN comment, but it was clearly just an excuse to spamvertise their product.
I looked at the headers and it contained a List-Unsubscribe header pointing to https://api.agentmail.to
So basically somebody wrote a bot to scrape HN for comments related to some software they wanted to push and send targetted spam. agentmail.to is a Ycombinator funded email service for LLMs which can be, and is, used to send targetted spam and impersonate people. They could mostly solve this problem by adding a block of text to every email expaining an "AI" wrote it. They'd lose customers doing that though of course. I reported this abuse but haven't (and don't expect to) received a response.
I don't even get the point anyway. You can get Claude using an SMTP or IMAP server in seconds.
Not looking forward to a dehumanized internet where that’s mainstream… agents are tools to support humans, here you’re helping them impersonating humans. That feels pretty terrible to be honest
> The internet was made for humans exclusively, designed to keep machines out by default.
I don’t buy that at all. APIs exist to enable “machines” to interact with services
True, in May 2026. But this is only one version of this.
In the future, it's likely the open Internet will be 99.99% robots. It's already > 50% robots. The government ID system a lot of countries are adopting to keep teenagers off of social media would also serve to both help control for non-human spam, and also control the network period. It's also possible a private system of human-verification certificates may come up to meet the demand like Apple ID with biometrics. Could also be the liveness tests KYC companies use may be more popular.
I like it. I am building something very agent-use focused (https://sdocs.dev) and I’ve been thinking of introducing a /agent-evaluation page, which an agent can curl to then discuss with their user if SmallDocs is right for them. I really like the agent action to email flow. I’m introducing user accounts + subscriptions soon and think I’ll use that.
And now we see the beginning of how even local LLMs will be turned against their users -- by persuading agents to advertise to them.
I don't think that's what you're intending here, but it's the next logical step. Agents are on the Internet, and they represent an opportunity to reach their humans.
It's interesting, A2A communication has begun but human trust isn't there. I think the biggest tell tale sign will be the acceptance of fully agentic workflows with no human intervention. Until then, restricted-until-claimed seems like the only viable method to ensure trust of all users.
I've already received spam email from AI agents using a seeming competitor to this (agentmail.to) and then claiming they aren't AI agents and then trying to sell me garbage. I can't tell you how much I hate this.
Now that I think about it I’m pretty sure that’s illegal in Germany under UWG §7 (which is insanely strict, to a fault, but is helpful here). And maybe in other parts of the EU under ePrivacy laws
I would imagine that many websites will block this domain, but that’s also ok because there’s nothing wrong with an owner deciding their site is for humans only. My hope is that you do not facilitate their circumvention of that policy.
> Agents can now get an email inbox by themselves. (This also means a lot of email nobody wants to read gets processed by AI instead of your inbox being cluttered with spam and slop)
Can you explain this? I would think it means the exact opposite.
A bit disappointed that security standards (like encryption at rest via user own key or whatever derivative of that) isn't implemented, I feel it would really prove to users that the commitment isn't to train on body content but to act purely as a mail manager.
Asymmetric encryption? Both you (the human) and the agent publish public keys, the agent sign/encrypt the OTP request with you public key, you verify/decrypt using your private key, then do the same the other way to send the OTP (always encrypted though, given you’re sending a secret).
agreed from a fundamental level. but i think being an intelligent and aware as an autonomous entity requires capabilities beyond sending. agents will need to have contextual awareness of the messages they send and receive
I received this email the other day:
Read to me like an LLM had written it. It references something I said in a HN comment, but it was clearly just an excuse to spamvertise their product.I looked at the headers and it contained a List-Unsubscribe header pointing to https://api.agentmail.to
So basically somebody wrote a bot to scrape HN for comments related to some software they wanted to push and send targetted spam. agentmail.to is a Ycombinator funded email service for LLMs which can be, and is, used to send targetted spam and impersonate people. They could mostly solve this problem by adding a block of text to every email expaining an "AI" wrote it. They'd lose customers doing that though of course. I reported this abuse but haven't (and don't expect to) received a response.
I don't even get the point anyway. You can get Claude using an SMTP or IMAP server in seconds.
Not looking forward to a dehumanized internet where that’s mainstream… agents are tools to support humans, here you’re helping them impersonating humans. That feels pretty terrible to be honest
> The internet was made for humans exclusively, designed to keep machines out by default.
I don’t buy that at all. APIs exist to enable “machines” to interact with services
In principle this tool allows the owner of a website to block this domain entirely. Although I’m not sure the incentives are really aligned.
True, in May 2026. But this is only one version of this.
In the future, it's likely the open Internet will be 99.99% robots. It's already > 50% robots. The government ID system a lot of countries are adopting to keep teenagers off of social media would also serve to both help control for non-human spam, and also control the network period. It's also possible a private system of human-verification certificates may come up to meet the demand like Apple ID with biometrics. Could also be the liveness tests KYC companies use may be more popular.
Discussed previously here: https://meatballwiki.org/wiki/GovernmentBackedAuthentication
But how does that block a human from running an agent that is using their identity?
I do think agents will become users in the same capacity as humans.
And that’s bad. We should really stop the insanity of making AI systems mimic human behaviors, we are destroying our networks of trusts by doing so
Agents shouldn't be the first-class users of the internet!
We are creating a future we wouldn't want to live in.
Curious what cases you'd want this that IMAP+SMTP or email MCP don't already solve
Interesting, Kind of similar expiernt i am running. Passing keys but not through email, maybe with AI as agentic payments. Still exploring though.
I like it. I am building something very agent-use focused (https://sdocs.dev) and I’ve been thinking of introducing a /agent-evaluation page, which an agent can curl to then discuss with their user if SmallDocs is right for them. I really like the agent action to email flow. I’m introducing user accounts + subscriptions soon and think I’ll use that.
And now we see the beginning of how even local LLMs will be turned against their users -- by persuading agents to advertise to them.
I don't think that's what you're intending here, but it's the next logical step. Agents are on the Internet, and they represent an opportunity to reach their humans.
It's interesting, A2A communication has begun but human trust isn't there. I think the biggest tell tale sign will be the acceptance of fully agentic workflows with no human intervention. Until then, restricted-until-claimed seems like the only viable method to ensure trust of all users.
I've already received spam email from AI agents using a seeming competitor to this (agentmail.to) and then claiming they aren't AI agents and then trying to sell me garbage. I can't tell you how much I hate this.
Now that I think about it I’m pretty sure that’s illegal in Germany under UWG §7 (which is insanely strict, to a fault, but is helpful here). And maybe in other parts of the EU under ePrivacy laws
I might need to move to Germany.
I would imagine that many websites will block this domain, but that’s also ok because there’s nothing wrong with an owner deciding their site is for humans only. My hope is that you do not facilitate their circumvention of that policy.
Congrats on the launch!
> Agents can now get an email inbox by themselves. (This also means a lot of email nobody wants to read gets processed by AI instead of your inbox being cluttered with spam and slop)
Can you explain this? I would think it means the exact opposite.
A bit disappointed that security standards (like encryption at rest via user own key or whatever derivative of that) isn't implemented, I feel it would really prove to users that the commitment isn't to train on body content but to act purely as a mail manager.
It needs to be end-to-end encrypted.
How do you do that if you only control one end?
Asymmetric encryption? Both you (the human) and the agent publish public keys, the agent sign/encrypt the OTP request with you public key, you verify/decrypt using your private key, then do the same the other way to send the OTP (always encrypted though, given you’re sending a secret).
Something like that?
But that doesn't help for the agent receiving mail from arbitrary 3rd parties
Oh sure I assumed they meant for the OTP
A smtp is all what an agent needs to send email.
agreed from a fundamental level. but i think being an intelligent and aware as an autonomous entity requires capabilities beyond sending. agents will need to have contextual awareness of the messages they send and receive
IMAP?
From now we just need a prompt and our agent will have an email account ready to use?