Tested runs of the go-to-market jobs people hand to Claude, ChatGPT and their own agents. Real prompts, real tool calls, real costs.
Ten rules any AI agent should follow before it emails a stranger on someone's behalf. Open, free to adopt, and written so a builder can check an agent against it line by line.
Seven tools that give Claude live prospecting data without an Apollo seat, with how each connects, how it charges and what it covers.
If you are building an agent that prospects, the API underneath decides what it can do and what every call costs. Eight APIs compared on the things that matter to builders.
Claude prospects as well as the tools you connect. Here are the best ones for each prospecting job, and the one that covers all of them in a single connection.
Claude can send an email from your own inbox, with your approval. That is a different job from sending a cold sequence to 200 people. Here is what each route does, on which plans, as of October 2026.
Claude Code is a good go-to-market workbench once five files are in place: the data connection, your context, your skills, your permission rules and a way to run it on a schedule. Here is each one, ready to copy.
Claude can run most of a sales team's outbound work, but only on top of the right infrastructure. Here is the stack, layer by layer: what each layer does, what to connect, and the three setups we'd start with.
Each play does one go-to-market job well and was run on live data before it was published. Install only the ones you need.
Clay is built around a table. Claude is built around a conversation. Connect Claude to live go-to-market data and the conversation becomes the workflow: you change it with a sentence, it uses judgement at every step, and it runs the whole loop.
A tested play for Claude, Claude Code, Cursor and Codex. Install it in one line, connect Layrcake once, and ask in plain English.
A tested play for Claude, Claude Code, Cursor and Codex. Install it in one line, connect Layrcake once, and ask in plain English.
A tested play for Claude, Claude Code, Cursor and Codex. Install it in one line, connect Layrcake once, and ask in plain English.
Claude can't look up an email address from memory, and it shouldn't guess one. Give it a finder and a checker, and it handles the list for you. Here is how, with the numbers from two of our runs.
Apollo is a sales platform with a large contact database and seats. Layrcake is infrastructure built for AI agents: one MCP server and one balance for finding, checking, researching and sending. Here is how they compare when an agent does the work.
Crustdata is a real-time B2B data API for agents. Layrcake is the go-to-market layer agents run outbound on: data, verified emails, signals, research and approved sends in one connection.
Instantly is a cold email platform with a lead database on a separate subscription. Layrcake gives an AI agent the whole loop from one balance: find, verify, research, then send only after a person approves.
ZoomInfo now offers a self-serve data API and MCP server for agents. Layrcake gives agents the data and the outbound loop that follows it: verification, signals, public posts, campaigns and replies, from one balance.
A tested play for Claude, Claude Code, Cursor and Codex. Install it in one line, connect Layrcake once, and ask in plain English.
A tested play for Claude, Claude Code, Cursor and Codex. Install it in one line, connect Layrcake once, and ask in plain English.
A tested play for Claude, Claude Code, Cursor and Codex. Install it in one line, connect Layrcake once, and ask in plain English.
A tested play for Claude, Claude Code, Cursor and Codex. Install it in one line, connect Layrcake once, and ask in plain English.
A tested play for Claude, Claude Code, Cursor and Codex. Install it in one line, connect Layrcake once, and ask in plain English.
A tested play for Claude, Claude Code, Cursor and Codex. Install it in one line, connect Layrcake once, and ask in plain English.
A tested play for Claude, Claude Code, Cursor and Codex. Install it in one line, connect Layrcake once, and ask in plain English.
Twenty official MCP servers for sales, grouped by the job they do, with how each one connects, what it needs to run, and when to pick it. Every fact from the vendor's own docs, checked 12 October 2026.
LinkedIn hasn't built a connector for Claude, but plenty of tools connect the two. Which one fits depends on the job: posting, outreach from your own account, reading public posts, or LinkedIn's own APIs.
Most Claude cold email sounds like Claude. Ours opened with a dated fact about each company, went only to checked addresses, and stopped at a draft. Here is the skill file, the run and the delivery rules.
ChatGPT can't look up who works where today. Connect it to a live sales data source and it can: find buying companies, brief you before calls, find verified emails and turn LinkedIn posts into leads.
We handed Claude six real prospecting jobs, one prompt each, and watched what it did. Here are the prompts, word for word, and what came back.
One skill file turns Claude Code into an SDR that never skips the boring parts. Type one line; get back a checked list, verified emails and an opener for every person.
We connected Claude to a live data layer and ran the whole loop on 20 real leads: find, verify, research, draft. Every tool call, and what each one cost.