Cold pitches, bounced
AI screens every stranger who writes to you. Pitches go to a pile you review when you feel like it, and real people always get through.
NewConnect Claude, ChatGPT or Cursor to your inbox
A fast shared inbox on your own domain. Give each AI agent its own address and decide what it may do. Plus the email API and MCP server for your product and your assistants. One place, one bill.
No card needed. Your domain is set up in about 10 minutes.
Illustration: the Koltrix inbox on a phone. New mail is sorted into Primary, a cold pitch is moved to Cold pitches, and a reply drafted by an agent waits until a person taps Approve and then Send.
IllustrationIn production
ReminderIt sends and receives its email on Koltrix
Your own domain
Mailboxes and team inbox on the domain you already own
Agents with inboxes
Give each AI agent its own address on your domain
7 days, no card
Start with a domain and real mail
yourcompany.com
SPFVerified
DKIMVerified
DMARCVerified
Draft, not sent
Hi Maya, yes, we can start the renewal on the 1st.
yourcompany.com
SPFVerified
DKIMVerified
DMARCVerified
1
Add the domain you already own. Koltrix writes the DNS records for you (ownership, MX, SPF, DKIM and DMARC) and checks each one live, so you know the moment mail can flow. About ten minutes.
Draft, not sent
Hi Maya, yes, we can start the renewal on the 1st.
2
Every new message gets a category and a one-line summary. Cold pitches go to their own pile. Threads that need an answer get a reply drafted in your tone, from the thread itself.
3
By default nothing leaves without a person clicking send. Approve a draft, change a word or skip it, and every send has an undo window. Auto, where an agent replies on its own when it is sure, is off until you turn it on for that agent.
Koltrix reads each new message once, gives it a category and a one-line summary, and files it. People writing to you always get through. Strangers selling to you meet Pitch Shield first.
Plenty of products now put AI in an inbox. Koltrix puts your team's inbox, your product's transactional email and your AI agents on one address book, one API and one bill.
AI screens every stranger who writes to you. Pitches go to a pile you review when you feel like it, and real people always get through.
Every “I'll send it Friday” is tracked, in both directions. You get a reminder before yours is due, and a polite nudge is drafted for theirs.
No more noreply. When someone replies to a receipt or a password reset, it lands in your team inbox threaded under the message that caused it.
Agent mail Beta
The same inbox your team uses, with guard rails. You see what every agent does in one list.
Auto replies on its own when it is sure: off by default per agent, capped, allowlisted and logged.
Each agent gets its own address and a key that opens that mailbox only.
Risky inbound mail is held back and you see why.
Agents draft or propose. A person sends. Every action is logged.
Support triage, invoices, leads and follow-ups, ready to install.
For developers
Send your product's email over REST or SMTP, hear about it with signed webhooks, and connect your own agents over MCP. Replies come back to your team, on the same domain.
Sending API and SMTP
Learn moreOne JSON POST or the relay. Idempotency keys and per-key permissions.
Signed webhooks
DocsHMAC-signed delivery, bounce, open and click events.
MCP server
Learn moreClaude, ChatGPT, Cursor and Claude Code over OAuth.
Agent mailboxes
Learn moreA scoped key per agent address, plus wait_for_email for sign-up tests.
OpenAPI spec
DocsThe whole public API in one file for client generators and agents.
curl https://api.koltrix.com/api/v2/emails \
-H "Authorization: Bearer kx_live_..." \
-H "Content-Type: application/json" \
-d '{
"from": "[email protected]",
"to": "[email protected]",
"subject": "Welcome!",
"html": "<p>Hi 👋</p>"
}'Replies to this message land in your team inbox, threaded under it.
New Works with Claude, ChatGPT and Cursor
Connect Koltrix to Claude, ChatGPT, Cursor or any app that speaks MCP, the open standard for giving AI assistants tools. You sign in and approve it once; it sees only the mailboxes you can, and an admin can switch it off for the whole workspace.
Paste this into your assistant. No API key.
https://mcp.koltrix.com/mcpWhat needs a reply from me today?
Used get_inbox_summary in Koltrix
2 conversations
Needs a reply · 14 unread: Primary 6, Other 5, Cold pitches 3
Draft a reply to Dana: yes, we ship Friday.
Used create_draft in Koltrix
Draft ready for your review
Claude and ChatGPT can read, search, organise and draft. Sending is off by default: an admin can turn it on, and the assistant must show you the message and get your yes before it sends.
Agents only send on their own if you turn on Auto for that agent. It is off by default, sends only when highly confident, to allowlisted recipients, inside a daily cap, and logs every send. One switch turns every agent back to Suggest. Agents have no tool to delete or forward mail, and you can switch AI off for a whole workspace.
Each workspace's mail is stored apart from every other customer's.
Agent actions are recorded, and you can switch AI off for a whole workspace.
Our AI providers are not allowed to train on your messages.
Agents cannot delete or forward mail. The capability does not exist.
“I built Koltrix because ReminderIt needed to send email and receive replies to it, and nothing I could buy did both. We are customer number one, which is a conflict of interest and also the reason the boring parts work.”
Every plan includes mailboxes on your own domain. Seats are included too, so adding a co-founder does not change the bill. Need more later? You can add extras after you start. Every plan starts with a 7-day free trial, no card.
The trial gives you 1 domain, 3 mailboxes and 50 sends a day for 7 days. See the full pricing and limits.
Yes. Add your domain and publish the five DNS records we generate for it: one that proves it's yours, MX, SPF, DKIM and DMARC. Koltrix goes live on it once they verify. The MX record is what moves your incoming mail to Koltrix, so publish it when you are ready to switch. If you already have an SPF record, we show you the merged version, so you never end up with two.
Only to do the job you asked for, and only while AI is switched on for your workspace. Sorting and spam scoring send the headers, subject and a truncated body to OpenRouter (TypeSafe's Jev model); summaries and drafts send the same to our generative provider, OpenRouter (Google Gemini and DeepSeek models). Neither is allowed to train on your mail. You can turn AI off per workspace, and then sorting falls back to our own spam engine and nothing leaves our servers.
Only if you turn it on. By default agents only propose, and a person clicks send. Auto is a setting per agent and is off by default. When it is on, the agent sends alone only if it is highly confident and can point to what its answer is based on, the recipients are on the allowed list and already known, the message it is answering looks safe, and the daily cap and any quiet hours allow it. Anything else waits for your tap in the Action center, and an identical message sent in the last 10 minutes is held too. Every autonomous send is logged, and one switch puts every agent back to Suggest. Agents have no tool to delete or forward mail.
Yes. Koltrix has an MCP server, so Claude, ChatGPT, Cursor and other MCP apps can search and read your mail, add labels, archive, and write drafts for you to review. You approve each app once with your Koltrix sign-in, and it only sees the mailboxes you can. Sending is off by default: a workspace admin can turn it on, you opt in when you connect, and the assistant has to show you each message and get your yes before it sends. It can never forward or delete mail. Setup steps are at https://docs.koltrix.com/mcp.
Koltrix is a web app today, and it installs to your phone's home screen. IMAP and SMTP access for desktop and mobile mail clients is on the roadmap for the release after launch, not in v1.
Nothing is lost. After 7 days without a plan your workspace becomes read-only: mail still arrives and your data stays safe and exportable, but you can open only 2 messages a day, and sending, the API and SMTP pause until you pick a plan. If it is still unpaid 30 days after that, we delete the data — after three warning emails.
In v1 we give you a written migration guide and a DNS cut-over checklist, and we will run the cut-over with you. There is no automatic mailbox importer yet; pulling your old mail across over IMAP is planned for the release after launch.
On servers we run ourselves in Frankfurt, Germany (European Union), kept separate from every other customer's, with attachments stored on our own storage. We are not built on another email provider's infrastructure.
Yes. Extra mailboxes are $1 each a month, or $9.50 each a year on annual billing. Starter includes 10 mailboxes and sells up to 27 more. Pro includes 50 mailboxes and sells up to 150 more. Starter with all of its extra mailboxes costs about the same as Pro, so past that point Pro is the better deal: more domains, seats and storage. Need more than 200? Contact us. The trial does not include extra mailboxes.
Yes. Every paid plan can add one storage package: +5 GB for $2 a month ($19.20 a year), +20 GB for $6 a month ($57.60 a year), +50 GB for $12 a month ($115.20 a year). It is billed with your plan and you can change or remove it from Settings, Billing. Starter with the biggest package still costs less than Pro, and Pro has more domains, seats and mailboxes, so move up when you need those too. The trial does not include extra storage.
An address on your own domain that belongs to an agent or a workflow, such as [email protected]. It has its own scoped API key that reaches that mailbox only, optional allowlists for who can reach the agent, and an audit log. Your team can open the same thread in the inbox. How many you can have depends on your plan.
Every inbound message is scanned when it arrives, with fast rules first and an AI check second, and gets a risk level of none, low or high with plain-language reasons. Whatever an agent reads is wrapped as untrusted data. For an agent mailbox you choose off, flag or quarantine, and quarantine is the default: a high-risk message is held back from the agent and shown to a person with a warning. No filter is perfect, which is why Auto is off by default and holds anything it is not sure about.
Every plan has a daily inbox send limit and a monthly API and SMTP limit, and they are enforced by the API, not just displayed. The exact numbers for each plan are on the pricing page and in the limits reference at https://docs.koltrix.com/limits.
No card required.