MoltSets MCP Server: Connect B2B Data to Your AI Agents

The MoltSets MCP server lets AI clients like Claude, Claude Code, ChatGPT, Cursor, Windsurf and Grok call MoltSets' B2B data tools directly: find work emails, carrier-verified mobiles, people and company search, all from a chat or an agent. You point the client at https://mcp.moltsets.com/mcp, sign in with OAuth credentials from the MoltSets app, and the tools appear. It is one of the more complete MCP setups for contact data, but a human must create the account, and every agent call draws on the same rolling Fair Use caps as the API.

Some tool links here are affiliate links (disclosure). To follow along, create a MoltSets account (free tier, no card). For pricing, data claims and weaknesses, read our MoltSets review.

Nerdy Joe is a certified Apollo.io Managed Services Partner and a PlusVibe Certified Silver Partner, and we are a MoltSets affiliate. We did not connect the MoltSets MCP server for this guide. The setup steps come from MoltSets' MCP, agents, CLI and library pages and its developer docs, checked November 9, 2026.

MoltSets MCP quick facts

What the MCP server is (and is not)

MCP (Model Context Protocol) is a standard way for AI clients to call external tools. MoltSets' version is a hosted server, so there is nothing to install or run locally: your AI client talks to MoltSets over HTTP and signs in through an OAuth consent screen.

MoltSets describes itself as "a go-to-market data platform for AI agents," accessed only through MCP or REST. Its FAQ is direct that there is no UI, no dialer and no sequencer. The MCP server is effectively the front end: your AI client becomes the place you search, enrich and export.

That shapes who it fits. If you already work in Claude, Claude Code or Cursor, it removes the need for a separate data dashboard. If your team wants a filterable list-building screen, MoltSets' own blog calls it "a poor fit."

Supported AI clients

These are the clients MoltSets lists on its MCP and integrations pages.

Codex appears on MoltSets' site alongside the CLI rather than as an MCP client. If you work in Codex, the CLI route below is the documented path.

MoltSets also lists platform and MCP partners such as Composio, Truto (20 MCP tools), Unified.to and others on its integrations page. Those are third-party routes we did not review.

How to connect MoltSets to an AI client

MoltSets documents the same three steps for every client.

1. Sign up and create credentials. Create a free account at app.moltsets.com/signup, open Connectors, and copy the Client ID and Client Secret.
2. Point the client at the server. Add https://mcp.moltsets.com/mcp as a remote MCP server or custom connector, with the OAuth credentials.
3. Approve the consent screen. Sign in on the MoltSets consent screen; the tools then appear in the client.

Then confirm the link works by asking the agent to call get_account. It is one of the free account tools, so the check costs nothing.

Claude Code

MoltSets documents this command:

claude mcp add --transport http moltsets https://mcp.moltsets.com/mcp --scope user --client-id YOUR_CLIENT_ID --client-secret

Claude Code then prompts for the secret and opens the consent screen. --scope user makes the server available across your projects rather than one folder.

Claude.ai and Claude Desktop

In Claude.ai, MoltSets says to go to Customize, then Connectors, and add a custom connector with the MCP URL and your OAuth credentials. On Team and Enterprise plans, an Owner adds it once for the organization. Claude Desktop is listed as supported too; MoltSets' page does not give separate steps for it.

ChatGPT

MoltSets lists ChatGPT support as a custom connector using Developer mode, on Plus, Team or Enterprise. Add the MCP URL and credentials as a custom connector, then approve the consent screen. MoltSets' page does not walk through ChatGPT's menus click by click, and ChatGPT's settings change often, so follow OpenAI's current connector docs for where the option lives.

Cursor, Windsurf, VS Code and Grok

MoltSets lists each as supported and gives the same server URL and OAuth details. The exact config screen differs by client and is not spelled out on MoltSets' page, so add a remote HTTP MCP server in that client's own settings using the URL, Client ID and Client Secret.

The setup prompt shortcut

MoltSets offers a second approach: paste a prompt asking your agent to set itself up. The prompt tells the agent to read https://moltsets.com/llms.txt, follow the Getting Started section, and confirm the connection with get_account.

This works well in agentic coding tools that can read URLs and edit config files. It does not remove the human step: you still create the account and the connector credentials yourself, because MoltSets says an agent can't create an account.

What your agent can do once connected

The MCP server exposes the same tools as the REST API. Our MoltSets API guide has the full reference; this is the short version.

Every email the agent returns carries an A to F send-safety grade and a last-validated date. Our MoltSets email validation guide explains how to tell an agent which grades to keep.

Prompts worth trying

These are our examples of the kind of request the tools support, not MoltSets output.

Use MoltSets to count how many VP or Head of Marketing contacts work at US software
companies with 51 to 200 employees. Use count_only first and tell me the number
before pulling any records.

For each LinkedIn URL in this list, get the best business email with MoltSets.
Keep only A and B risk scores, put C grades in a separate table, drop D grades,
and give me a CSV.

Call get_usage for the week and tell me how many enrichment records I have left
before you start this job.

The pattern to copy is in the last prompt: make the agent check headroom (free) and size the job before it spends records. MoltSets also notes that search_linkedin_profile with count_only=true returns counts without charging, which is a cheap way to size a segment.

The CLI and Claude skills

MoltSets CLI

For bulk CSV work, MoltSets' CLI is often better than a chat. Run it with npx moltsets --help or install it with npm install -g moltsets (package @moltsets/cli, Node.js 18 or later).

moltsets enrich file.csv --column linkedin --out enriched.csv

MoltSets documents --dry-run (counts rows and checks headroom with no paid calls), --limit 10 for a sample, batches of 100 with progress saved after each batch, automatic waits for Fair Use resets, and up to four retries on a 429. The output keeps your original columns and adds emails, risk scores and a status per row.

The CLI authenticates with an API key (flag, MOLTSETS_API_KEY environment variable, or moltsets login), not the MCP credentials. It also has moltsets skill install, which installs an agent skill, so a coding agent can drive the CLI for you.

GTM Skills Library

MoltSets publishes 82 pre-built Claude Skills (its count) in its library. Examples it lists include:

• HubSpot Email Backfill
• HubSpot Mobile Phone Backfill
• HubSpot Buying Committee Expansion
• Build an Outbound Prospect List
• Enrich RB2B Visitors
• IP to company

Skills package a workflow (which tools to call, in what order, with what checks) so you do not have to write the prompt each time. Read any skill before you run it on a live CRM, since a backfill skill writes data.

Honest limits for agent use

The MCP server makes MoltSets easy to call. It does not change what the account can do.

• A human signs up. MoltSets says an agent can't create an account. Someone has to own the account, credentials and billing.
• Same Fair Use caps. MCP calls share the account's limits with every API key and connector. On Solo ($27) that is 1,000 enrichment records per 5 hours and 5,000 per week; search is 500 per 5 hours and 2,500 per week, as of November 2026.
• Misses still count as requests. Records are deducted only on a hit, but every lookup counts toward the request cap (5x the record cap). An agent looping on a poor list can burn that.
• 25 results per search call. search_people pages at a max of 25, with 3 to 8 concurrent searches by plan.
• Mobiles cost extra. $0.15 per number returned on a separate monthly phone token subscription, paid plans only. MoltSets does not return Do Not Call status, so DNC scrubbing and TCPA consent are on you.
• Personal emails are US only, and capped on Team and Max.
• Ad audience tools are not on Team or Max since September 29, 2026.
• Validation and scraping are not live. validate_emails and scrape_website are priced but marked coming soon.
• No published accuracy figure, database size or bounce guarantee. MoltSets' own accuracy blog says to test it on your data like any other provider.
• Young product. Tool names, limits and response behavior changed several times in September and October 2026, per its changelog, and it has few or no public reviews yet.

Our MoltSets review covers these weaknesses alongside the pricing. Our practical advice: give agents a budget in the prompt ("stop after 500 records"), require a get_usage check first, and keep D-grade emails out of anything that reaches a sequencer.

How MoltSets MCP compares with other data MCPs

MCP support is spreading across data vendors. This is what each vendor's own pricing page showed when checked in November 2026.

The main difference is the model. MoltSets charges a flat fee with rate windows, so an agent's cost is predictable but its volume is capped; credit-based vendors charge per result, with no weekly ceiling beyond your balance. Findymail adds a bounce guarantee and native sequencer pushes, which MoltSets does not have; see our Findymail review.

If your agents need a UI fallback for humans, Apollo combines a searchable database with sequencing (we are an Apollo.io Managed Services Partner); our Apollo review covers it. Our best B2B data providers roundup compares the rest.

Who the MoltSets MCP server fits

• Best for: founders, RevOps and agency operators who already work in Claude, Claude Code or Cursor and want contact data inside that workflow at a low flat price.
• Best for: builders wiring B2B data into custom agents, where the same tools are also available over REST.
• Not ideal for: sales teams that want to click through filters in a dashboard, or that need a sequencer and dialer in the same tool.
• Not ideal for: very high or bursty volume beyond your plan's weekly cap.

Final take

The MoltSets MCP server is a straightforward way to give Claude, ChatGPT, Cursor and other AI clients real B2B contact data: one URL, OAuth credentials, and the full tool set, plus a CLI and 82 Claude skills for bulk and repeatable jobs. The limits are the account's limits: human sign-up, weekly Fair Use caps, paid mobiles and no published accuracy figures.

The free tier is enough to test it in your own client. Sign up for MoltSets, create connector credentials and run get_account, then try a small enrichment before you move up to a paid MoltSets plan.

For the full picture on pricing and data, read the complete MoltSets review, and if you want the step-by-step basics first, see how to use MoltSets. When you are ready, connect MoltSets to your agent.

Rather not build the list yourself?

Nerdy Joe is owned by Agency Growth Partners (AGP). AGP builds verified lead lists and runs outbound for agencies, including agent-driven enrichment like the setup above. If you would rather hand it off, book a call with David.

Sources

• MoltSets MCP server: https://moltsets.com/how-it-works/mcp-server/
• MoltSets for AI agents: https://moltsets.com/how-it-works/agents/
• MoltSets developer docs index (llms.txt): https://developer.moltsets.com/llms.txt
• MoltSets llms.txt: https://moltsets.com/llms.txt
• MoltSets API page: https://moltsets.com/how-it-works/api/
• MoltSets rate limits: https://developer.moltsets.com/getting-started/rate-limits
• MoltSets integrations: https://moltsets.com/how-it-works/integrations/
• MoltSets CLI: https://moltsets.com/how-it-works/cli/
• MoltSets GTM Skills Library: https://moltsets.com/library/
• MoltSets FAQ: https://moltsets.com/faq/
• MoltSets pricing: https://moltsets.com/pricing/
• MoltSets mobile numbers pricing: https://moltsets.com/pricing/mobile-numbers/
• MoltSets email risk scores: https://developer.moltsets.com/moltsets-data/email-risk-scores
• MoltSets blog, best B2B data enrichment APIs: https://moltsets.com/blog/best-b2b-data-enrichment-apis/
• MoltSets blog, how to measure contact data accuracy: https://moltsets.com/blog/how-to-measure-contact-data-accuracy/
• MoltSets changelog: https://support.moltsets.com/en/articles/15424421-changelog
• Findymail pricing: https://www.findymail.com/pricing/
• Lusha pricing: https://www.lusha.com/pricing/
• Cognism pricing: https://www.cognism.com/pricing
• Apollo.io pricing: https://www.apollo.io/pricing
• ZoomInfo pricing: https://www.zoominfo.com/pricing
• People Data Labs pricing: https://www.peopledatalabs.com/pricing