Principle I
Outreach has six variables, not one
Picture a Rubik's cube. Six faces. You twist all six together — or you don't get there.
- I
Who
The right company
- II
To whom
The right person
- III
Where
The right channel — email, LinkedIn, WhatsApp, Telegram, X, Facebook
- IV
When
The right time — when they have a minute
- V
What
The right context — about them, not about you
- VI
Why
The right offer — they actually need it
Break the balance of one face — and you're spam. Volume doesn't make spam. Imbalance does.
Principle II
Email is 20%. The spectrum is wider.
About a fifth of replies come through email. The rest comes from wherever your customer already shows up.
Two ways to send
Fast start
Branded free mailboxes — name.company@gmail.com
You want to start now. In 2026 these mailboxes pass spam filters easily. Isolated — easy to rotate.
Slow start
Your own brand domain — name@brand.com
Ready to invest in infrastructure and precise targeting. Enterprise filters are harsh. Warmup is a marketing pitch — providers actively fight it.
Links and pixels — our recommendation
Don't use any: no redirect links, no tracking pixel, no image links. Modern anti-spam systems analyze images and detect the pattern. Use links only when you're 100% sure you're not spam — and have built reputation.
Channels beyond email
Your customer has dozens of entry points. Use the ones that fit the case:
- 01
General inbox info@company.com — normal practice, often faster
- 02
Contact form on the website — shorter path to the right person
- 03
Messengers — WhatsApp, Telegram
- 04
Social — LinkedIn, X, Facebook
- 05
Form-fill widgets — automate, but keep messages unique
Principle III
Creative has no recipe
AI gives you the base. The salesperson knows the customer and the product better. There are no ready recipes — you find the right note in the process. The note that resonates with this specific person.
AI is the tool. The core is you.
Protocol
Seven steps. On a loop.
Take a step — move to the next. Reach the seventh — return to the first.
- 1
Find companies
There are catalog services for this. Ask ChatGPT — it'll suggest a few. Pick the ones that match your market.
- 2
Build a dossier
On the company. On the person. Look for a signal, a hook, a real reason to write — a specific trigger to message this person right now. No reason — don't write. Strictly: only people who actually need it. Don't bet on "100 sent, maybe 1 lands" — spam filters react fast to the text itself and pull even previously-delivered messages out of the inbox. Then ask ChatGPT: "Does this fit my ICP? Is there a real reason to reach out?"
- 3
Describe your ICP
ICP — Ideal Customer Profile. You're a salesperson, you know your customer better than anyone. Describe them and pass it to AI.
- 4
Ask AI about timing
Claude or ChatGPT will suggest the best moment to write to this specific person.
- 5
Build a sequence
First message. Two days later — a different channel or a follow-up. Build the sequence with AI. One person at a time, no template.
- 6
Send until first spam hit
Watch inbox placement in real time. The moment you start landing in spam — change the creative, the angle, and the mailbox. Don't push through. Check inbox placement
- 7
Loop it
Lists → dossiers → messages → lists. Warm up with social signals — likes, follows on LinkedIn — but inside platform limits. Don't be spammy.
Three ways to apply
Pick your level of automation
Same method. Different effort.
DIY
Run the protocol by hand
All tools are public: catalogs, ChatGPT, your own mailboxes. Slower, but you're in full control.
With LDM
AI spins the cube for you
Connect to LDM. The platform handles dossiers, sequences, channels, inbox monitoring, rotation. You set the direction.
Agent mode
Delegate to an agent
You give the task to an AI agent (Claude Code, ChatGPT, Cursor, Manus). The agent connects to LDM via MCP/A2A and delegates delivery to us — deliverability, rotation, and infra are what the agent can't pull off on its own.
The platform
The protocol, running in LDM
The Live Direct Marketing philosophy is simple: the platform is a transparent conveyor for a protocol you've already validated — not a spam cannon. Every email is verified for actually landing in the Inbox: Inbox is billed, Spam and non-delivery are not. Our incentive is the same as yours — delivered emails to real people. Here's how the seven steps map onto the platform:
01 · List
Steps 1–3 — companies, dossiers, ICP. Import from CSV or via API; the platform de-duplicates and runs every address through stop lists and suppression — unsubscribed addresses physically can't be mailed.
02 · Creative
Steps 4–5 — timing and sequence. Build the email with variables and spintax, or generate AI variants per segment. Run it in the sandbox on your own contacts and send a test to yourself first.
03 · Campaign
Step 6 — send with live inbox monitoring. Pour the audience, check readiness — accounts, limits, send window — and confirm the launch. Nothing goes anywhere without an explicit approval.
04 · Dialogs
Step 7 — the loop. Replies land in a single inbox: classified as interested / not now / refusal, answered right from the platform, and an interested contact becomes a lead in the pipeline.
The platform automates the protocol you already walked by hand. It doesn't replace the search — it makes it cheaper to repeat.
MCP
Your AI agent is a full LDM operator
LDM was designed for AI agents from day one. MCP (Model Context Protocol) is an open standard that lets Claude Desktop, Cursor, and any compatible client connect to the platform as a set of tools. Everything you can do in the UI, an agent can do over MCP: lists, creatives, test sends, campaigns, dialogs.
The key difference: the platform answers the agent not as a bare API but as a market expert. Every response carries an _expert block — what to do next, the typical flow, known pitfalls. The capability map and the recommended end-to-end scenario live at /api/v1/agent-guide — the agent reads them itself; you don't have to write it a manual.
Launch in three steps
- 01
Issue an API key in Settings. One ldm_* key works for REST, MCP, and A2A at once.
- 02
Connect the MCP server in your client — Claude Desktop, Cursor, or any other. Config below.
- 03
Give the agent a task in plain words: “Import companies from CSV, build a list, prepare a creative with variables, send a test to my address, and show campaign readiness.” It walks the protocol itself, guided by the _expert hints.
{
"mcpServers": {
"ldm": {
"command": "npx",
"args": ["-y", "--package=@live-direct-marketing/sdk@latest", "ldm-mcp"],
"env": { "LDM_API_KEY": "ldm_..." }
}
}
}The agent prepares — the human decides: launching a campaign requires explicit confirmation, and every agent action is visible in the audit log.
MCP is not a side feature. It is the platform's primary interface: the human sets the goal, the agent runs the protocol.
Legal
White-hat only
Outreach only survives long-term when it's lawful. These rules are built into the platform — you don't have to remember them, and you can't bypass them:
- 01
An honest sender: real name, real domain, no deceptive subject lines. No masks, no spoofing.
- 02
Every email carries a working unsubscribe link. Once someone opts out, the address goes to suppression — the platform itself won't let you mail it again.
- 03
Stop lists and suppression are checked automatically before every single send — not “when convenient”.
- 04
Write to business addresses with a legitimate business reason. Handle personal data strictly by the law (GDPR and local equivalents): minimal data, deletion on first request.
- 05
Purchased grey lists and attempts to dodge unsubscribes are a direct violation of the platform rules and grounds for a ban.
Domain reputation and legal cleanliness are the same asset. Spent in a day, earned over months.
For your case
Pros and cons aren't public
They depend on your industry, ICP, stage, and channel. Write to us — we'll review your case and suggest where to start.
Simple market logic: if a competitor found a working process, they won't tell you. They'll only tell you once it stops working for them. Everything else is info-marketing without real proof in practice.
Any questions?
Write to us — we will reply personally. No funnels, no forms.
Is this an LDM ad?
No. We are not selling LDM to you. There are dozens of solutions on the market — ask AI and pick what fits your case.
Why this article then?
We're sharing what we've learned, so you don't end up a hostage of "info-marketing".
What if my emails go to spam?
We're against spam. An email in spam means you're doing something wrong. Treat it as a signal, not as a reason to push harder.
And AI?
Value human conversation — that's what closes deals. Let AI be the helper, not the replacement.
From Artyom, the developer: I value networking and plain human conversation — voice calls, messengers, however works. No LDM pitch, no agenda. Just glad to connect.