The Agent-Run Inbox
One paste. Five questions. Then it operates.
Cold email infrastructure, installed into your AI agent.
The Agent-Run Inbox installs a cold-email operating system into Muse, ChatGPT, or Claude. Your agent sets up the infrastructure, gradually warms the sending identity, monitors sender health, and pauses itself when something goes wrong. You approve. It operates.
One-time purchase. Google Workspace only in V1.
You
Paste INSTALL.md
Your AI agent
I need access to your DNS provider to complete this automatically. Connect it and I'll make and verify the changes. Or I can give you the exact manual steps.
System
- Mission 3
- DNS
- Status
- In progress
- Proof
- MX / SPF / DKIM / DMARC verified
How it works
Three steps. No system to learn.
Download
You get INSTALL.md, the file your agent runs. One file is the whole product.
Paste it into your agent
Use Muse, ChatGPT, or Claude. Answer up to five questions. That is the entire setup.
Your agent starts operating
It works through seven missions in order: diagnose your domains, pick a safe sending domain, configure DNS, verify the sender identity, warm it gradually, cut over to cold outreach on evidence, and keep operating it every day. You approve the human steps. It does everything else and reports back short.
On your phone, in a plain chat app, the agent hands you short exact checklists for anything it cannot click itself, then verifies each result you paste back. The proof gates are the same either way.
Connection-first
Connect it or don't. The proof stays the same.
Every mission starts the same way. The agent asks to connect the account it needs and tells you in one sentence what access it wants and what that unlocks. Connect it, and the agent does the work itself and verifies the result. Prefer not to connect it? The agent gives you the shortest exact manual instructions, you do the clicks, and it verifies afterward. It never assumes a connection exists. It never claims it did something it did not do or verify.
You choose what to connect. Every connection is narrow: one mission, least privilege, revocable when the phase ends. Nothing connects without your explicit say-so.
Path A
Connect account
Agent performs the action, verifies it, then continues.
Path B
Don't connect
Agent gives exact manual steps. You perform them. The agent verifies, then continues.
Same proof gate. Nothing advances on claims.
Built to operate
Less learning. More verified work.
One paste, five questions.
No folder to understand, no deliverability degree required. The agent starts doing real work in minutes.
Proof before progress.
Every phase ends with observable evidence, and cutover happens on clean history rather than a calendar date. Attempted work is never treated as proof.
It stays on after launch.
The agent keeps watching bounces, complaints, and reputation, and pauses itself before a dip becomes damage. It never scales through a dip.
Your rules, not a course.
Per-send approval, auto-send once warmed, and reporting cadence are your call. The strict defaults are safe; you can relax any of them.
What you receive
The install, the persistent system, and the manual.
The files are yours. There is no subscription, dashboard, or account with us.
INSTALL.md The operating system itself. Paste it into your agent and it runs. This is the product.
The file set for agents that keep files across sessions: your config, saved state, mission checklist, reporting template, and troubleshooting notes. Your agent maintains these so it remembers where it left off.
The Agent-Run Inbox manual (PDF) The human-readable reference for what is happening and why. Read it for understanding, not for execution.
All V1 updates Updates to this Google Workspace version are included.
Fit
For operators who want the system run properly.
It is for
- Founders and operators who want cold email done properly and would rather have their AI agent run the infrastructure than learn deliverability themselves. You have Google Workspace, or you are willing to get it.
It is not for
- Anyone on Microsoft 365. V1 is Google Workspace only, and it says so plainly.
- Anyone looking for a course to study, a prompt pack, or a list of tricks. There is nothing to study. The agent does the work.
- Anyone who wants email sent with zero involvement. You approve. The agent operates. Those are the terms.
FAQ
Straight answers before you install.
Does this send cold email for me?
It builds and operates the sending system: the domain, the authentication, the warmed identity, the daily sending loop, and the monitoring. You choose the recipients and approve what goes out. It does not write your strategy.
Does it require a warmup service?
No. No dedicated warmup software is required or used. Warmup is real, human-style correspondence the agent queues and delivery-checks.
Does it work with Microsoft 365?
No. V1 is Google Workspace only, stated up front. It will not invent a Microsoft workflow.
What if I don't want to connect my DNS or Workspace account?
Then don't. The agent gives you the shortest exact manual instructions, you do the clicks, and it verifies the result afterward. Nothing is assumed.
What does the agent actually do?
Seven missions in order: diagnose your domains, pick a safe sending domain, configure DNS, create and verify the sender identity, run the warmup queue, cut over to cold outreach only on clean evidence, and operate the system daily with automatic pauses on trouble.
Will this run automatically every day?
If your agent supports scheduled tasks, the daily cycle can run on its own. In a plain chat, you return and say: Run today's Agent-Run Inbox cycle. Either way it resumes from saved state.
Is this SaaS?
No. No subscription, no dashboard, no account with us. Files you own, running inside your own AI agent. $29 once.
What if my existing domains are unsuitable?
The agent tells you plainly and stops rather than forcing a bad choice. The smallest safe next step is registering a clean, neutral domain, typically $10-15/year.
Do I need to understand SPF, DKIM, or DMARC?
No. The agent configures and verifies them and refuses to advance until they check out. The PDF manual explains what is happening if you are curious.
What AI agents work with it?
Any capable agent: Muse, ChatGPT, Claude, or similar. Agents with file access and scheduling get the full persistent experience. Plain chat agents keep state in the conversation and work the same way.
$29 once
Install it into your agent.
One file turns a capable AI agent into a careful outbound infrastructure operator.
V1 operates with Google Workspace only. No dedicated warmup software is required. Depending on your setup, you may need to register a clean sending domain (typically $10-15/year).
Already purchased? Open the install page