MCP — Model Context Protocol — is an open standard for connecting AI assistants to outside systems. Every system that implements it becomes a set of tools with real parameters and real results, not a web page the assistant has to be told about.
Point an MCP-capable tool at https://app.leadspicker.com/mcp, authenticate once against your own Leadspicker account, and the platform shows up as callable tools: search the lead database, read and build lists, enrich contacts, prepare outreach sequences, check what something will cost in credits. Your AI tool stays the interface, MCP is the bridge, and the database, enrichment and outreach engine are what it exposes.
companies in the lead database
people, by role and seniority
one currency across everything
Database size from leadspicker.com. Credit pricing: Subscription → How Credits Work in the app.
The assistant handles the language. Leadspicker handles the data and the work.
When you ask your assistant to find people, enrich a list or draft a sequence, it routes the request through the Leadspicker MCP server. Leadspicker runs it against the same backend the app uses — the same database, the same enrichment providers, the same credit balance — and returns structured results into the conversation.
Authentication is OAuth against your own account. There is no API key to generate, copy or rotate: your CLI logs in the way you log into the app, and the server only ever reads and changes your own workspace.
Leadspicker MCP is a remote server, so there is nothing to run locally and nothing to keep updated. Anything that speaks MCP can connect to it.
Three CLIs have one-line setup. Or run the installer: it finds every supported CLI on your machine, registers the server and the assistant skills in each one, and starts the login flow. Any other MCP client — Cursor among them — connects to the same URL through its own “add a remote MCP server” flow. The server URL and the commands are also in the app under Integrations → MCP.
Four jobs that normally span a dozen tabs. Each is something you ask for in plain language, and the context carries from one to the next.
Search 70M+ companies and 500M+ people by industry, size, revenue, location, role and seniority — or describe the target in your own words and find look-alikes of a company you already win with. Preview the count before you commit to anything.
Verified emails, phone numbers, technographics, hiring data, funding. Waterfall enrichment runs across providers until something answers, and you can bound a run to the first N rows while you are still testing the shape of a list.
Create a list, import the matching records into it, draft a multi-step sequence. Drafts stay drafts until a human opens the app.
Credits are one currency across everything: one credit is one euro cent. The assistant can price an enrichment run before you start it and tell you where the balance went afterwards.
Same work, same data, same credits. What changes is how many windows it takes.
One conversation, context carried between steps
CRM, data tool, LinkedIn, company site, spreadsheet, email tool
You do not have to adopt Leadspicker as your entire go-to-market system. Take exactly the slice you need — the data, the enrichment or the sending — and orchestrate it inside a pipeline you build yourself.
This is where Leadspicker MCP differs from a connector that simply mirrors a product into a chat window. The MCP server does not care which parts you call or what you wrap around them — in your own code, alongside your own tools. Leadspicker can be a step in your workflow rather than the workflow.
Installing the MCP installs a set of skills — written procedures the assistant loads when the task matches, so it already knows the right order of operations and the house rules. A new person gets the vetted path on day one, and there is a skill for authoring your own.
Searches by similarity when you name a reference company, instead of guessing industry filters.
The data model everything depends on: a list holds contacts, a sequence holds steps, and contacts reach a sequence only by connecting a list.
Which enrichment columns to run, and in what order.
AI columns for scoring, labelling and writing opening lines.
Email, LinkedIn and multichannel sequences with branching.
Scheduled scrapers that keep topping a list up.
Checks what is actually connected before answering anything about a CRM.
Buying signals, financials, hiring, funding and technology.
MCP is for in-the-moment, in-your-flow work. The app is for volume, for oversight, and for anything you want to look at rather than read back.
The connector reads and changes only your own workspace, and the public tool surface deliberately cannot send email or LinkedIn messages, cannot activate a sequence and cannot change account settings. Drafts are drafts until a human opens the app. Imports are the one thing that spends your allowance, so they are flagged as such and your assistant should ask before running one.
Everything above is Leadspicker as an MCP server. It also works the other way: connect any MCP server to Leadspicker and use it as a data source.
If a provider you need is not built in, you bring it. If the data is not public at all, you can still target over it — your own product usage data, your own account history, your own internal systems, not only the external company data your competitors can also buy. Pair it with a custom HTTPS request for anything that has a plain API, and the set of things you can enrich or segment against stops being a fixed list.
Connect one tool and run one task end to end. Pick a real account you want to reach and do the whole thing in one conversation.
Run curl -fsSL https://app.leadspicker.com/install.sh | bash — or add the server to your CLI by hand.
Log in when the OAuth flow opens. It is the same login you use for the app.
Ask for companies that look like one of your best customers.
Ask for the right contact at each company, with a verified email.
Ask what the enrichment cost, then let the assistant draft the sequence.
Read the draft in the app. If you type the same request every week, write it down as a skill — that is where an ad-hoc prompt becomes a standard.
Didn’t find your answer? We are happy to walk you through it on a call.
A remote MCP server at https://app.leadspicker.com/mcp that exposes the Leadspicker platform — lead database, enrichment, lists, sequences, credits — as tools your AI assistant can call. MCP is an open standard, so any assistant that supports it can connect.
Claude Code, Codex and OpenCode have one-line setup and are configured automatically by the installer. Any other MCP-capable client, Cursor included, connects to the same URL through its own remote-server flow.
No. Your CLI authenticates against your Leadspicker account over OAuth, the same login you use for the app. Nothing to generate, store or rotate.
No. Work driven over MCP costs exactly what it costs in the app, out of the same credit balance — one credit is one euro cent. The current price list lives in the app under Subscription → How Credits Work, and the assistant can price a run before you start it.
No. The public tool surface cannot send email or LinkedIn messages, cannot activate a sequence and cannot change account settings. It finds, enriches, builds lists and prepares drafts; a human opens the app to launch.
The API is for systems you build and maintain. MCP is for the assistant you are already talking to — no client to write, no schema to keep in sync, and the skills mean it already knows how Leadspicker expects things to be done.
Yes, that is the second direction. Any MCP server can be connected into Leadspicker as a data source, alongside custom HTTPS requests for anything with an API — so you can target over internal data that is not available publicly.
