Lusha MCP review: 22 tools, two integration paths, one shared rate limit

By Jan Suski. I go deep on every review, checking every number against its source and digging up details other coverage misses, a habit from six years of SEO and link building for tech companies. Bylines at Surfer, Brand24, and Senuto. How the scoring works.

Published 11 September 2026Prefer this source on Google

Sales intelligence

Lusha runs two different things under the same name, and mixing them up is easy to do by accident. One is a two-tool open-source package on GitHub that a developer runs themselves. The other is a hosted, 22-tool server that shows up directly inside Claude’s and ChatGPT’s own connector directories.

This review breaks down what ships through each, what the REST API can do that neither covers, and where Lusha’s own marketing pages disagree with each other on access and on the accuracy of the data underneath it.

MCP, short for Model Context Protocol, is the open standard that lets an AI assistant call a vendor’s own tools mid-conversation and pull real data into the chat. For a sales team, that means asking Claude to pull a list of VPs of Sales at Series B companies and getting back real, credit-metered contact records grounded in Lusha’s own database.

This review is scoped to the MCP layer specifically: what it exposes, what it costs, and how it holds up against Lusha’s own REST API and against three competitors running the same play. For contact accuracy, the credit system, and the rest of the product, the site’s main Lusha review covers the full picture; this page exists because the MCP server is now big enough, and inconsistent enough across Lusha’s own pages, to need its own accounting.

Anyone already paying for Lusha and weighing the MCP connector’s ten-minute setup against its payoff, or evaluating Lusha partly on how well it plays with an AI-agent workflow, is the intended reader here.

↑ FACT SHEETTwo servers, one name

Lusha publishes an open-source MCP server on GitHub, lusha-oss/lusha-public-api-mcp: a two-tool npm package (personBulkLookup, companyBulkLookup) that authenticates with a Lusha API key passed as the LUSHA_API_KEY environment variable and caps a bulk request at 100 contacts or companies. As of this review it carries 6 stars and 2 forks, Apache-2.0 licensed.

The server that shows up in Claude’s and ChatGPT’s own connector directories is a different, hosted one at mcp.lusha.com. It exposes 22 tools across contact and company search, enrichment, prospecting filters, signal lookups, lookalikes and account usage, authenticated over OAuth for Claude, ChatGPT and Codex, or an API key for VS Code, Cursor, n8n and Gemini CLI.

Scalekit’s independent comparison of the two access paths counts the same 22 tools on the hosted side.

Neither server replaces the other. The open-source one suits a developer who wants a small, self-hosted tool with full control over the code. The hosted one is what a rep connects to from inside a chat window, and it’s the one every figure in this review refers to unless stated otherwise.

↑ FACT SHEETSetting it up and the first prompt

Setup runs through claude.ai’s connector directory or the matching ChatGPT app entry: add the connector, complete OAuth, confirm the tools show up in the chat’s tool menu. Lusha’s own walkthrough demonstrates it with a real prompt: find Series B SaaS companies in the UK with 50 to 200 employees that hired a VP of Sales in the last 90 days, followed by a second prompt pulling verified direct dials and work emails for the VP of Revenue at each result.

A sponsored LinkedIn post by GTM consultant Maja Voje, disclosed as paid but built around a workflow she says was implemented for a real client, walks through a similar pattern with Lusha and Claude’s MCP connector: build a contact list of 2,000-plus marketing agencies in Texas, enrich it with Lusha to surface 25 new leads a day, then use the connector to draft outreach emails without leaving Claude.

It’s sponsored, promotional content, and it should be read with that label attached. The workflow it describes still lines up with what Lusha’s own documentation says the connector does.

No independent account beyond that has surfaced. A search across Reddit, Hacker News, X and G2 for commentary specific to Lusha’s MCP server turned up nothing at the time of this review. The connector is new enough that outside opinion of it hasn’t formed yet.

↑ FACT SHEETWhat the REST API can do that MCP can’t

A single MCP call tops out at 25 companies or 50 contacts, against 100 records on the REST API, sized to what Scalekit’s write-up calls a 25,000-token response budget. Four things the REST API can do are missing from MCP altogether: webhooks, the Tables feature, Decision Makers, and Website Visitors matching.

MCP calls draw from the same credit pool as the REST API, and both authenticate through the same account. A failed lookup isn’t charged, but one rep running a heavy prospecting session through Claude can eat into the rate limit the rest of the team’s API calls share, a detail neither Lusha’s marketing pages nor its webinar recap mention.

Scalekit’s own conclusion, from a company that sells credential management and doesn’t compete head-on with Lusha, is that MCP suits interactive research on modest volumes while the REST API stays necessary for bulk enrichment, webhooks and anything that has to run on a schedule. Production teams, in its account, typically run both at once.

↑ FACT SHEETWho can turn it on

Lusha’s own account of who gets access has moved. An August webinar recap said MCP required a Sales Starter plan or higher. Its current documentation says any plan qualifies, and the setting is on by default in account settings, no separate purchase.

That leaves a real question for anyone reading an older guide or a webinar recording: the account settings are the one reliable answer, since Lusha’s own public pages have already stated two different ones.

↑ FACT SHEETThe database Lusha markets through MCP

The MCP landing page markets two data layers: a search layer with 290 million-plus contacts, 195 million-plus emails, 117 million-plus direct dials and 26 million-plus companies, and an intelligence layer built on 7 predictive models, 24 buying signals and 10,000-plus lookalike profiles. The company count doesn’t hold still within the same page: the hero subheading above those boxes states 29 million-plus companies, three million more than the stat box beneath it.

Those numbers also don’t match the figures Lusha publishes elsewhere. The main Lusha review documents a general database of over 280 million direct dials, more than 152 million business emails and over 30 million company profiles, sourced independently. The MCP page’s 117 million direct dials is a materially smaller number than the 280 million figure quoted on Lusha’s general marketing, for what reads like the same underlying database.

Accuracy claims follow the same pattern. The MCP landing page cites roughly 90% overall; the Claude integration post narrows that to 85% phone and 97% email, for European markets specifically. Neither matches the 98% email and 86% phone Lusha publishes on its main pricing page, the same claimed-versus-reported gap the main review’s data-quality section already tracks with real campaign results.

↑ FACT SHEETLusha MCP alternatives

ZoomInfo runs its own MCP connector across six core functions: finding and enriching accounts, finding and enriching contacts, and researching both, live for Claude, ChatGPT, Perplexity and Replit. New ChatGPT connector sign-ups get 100 free credits to test it. A team already choosing between ZoomInfo and Lusha on data depth gets the same MCP-versus-REST tradeoff on either side. See the Lusha vs. ZoomInfo comparison for the rest of the platform-level differences.

Apollo.io’s MCP connector works on any plan including Free, the same access model Lusha eventually landed on, reaching Claude, ChatGPT and Perplexity against a claimed 230 million-plus verified contacts. One customer Apollo cites enrolled 530-plus contacts into sequences across five industry verticals in three months, work it says previously took 15 to 20 hours done by hand.

Apollo.io bundles that data layer with outreach sequencing in the same platform, a combination Lusha doesn’t offer on its own.

Seamless.AI shipped a native MCP server on 6 May 2026, also reaching Claude and ChatGPT, for finding and enriching contacts, building lists and researching accounts through natural-language prompts. Its own base-platform review and dedicated MCP writeup document real deliverability complaints against Seamless’s “verified” label, the same category of complaint that shows up in Lusha’s own accuracy discrepancies above.

FullEnrich runs its MCP server at mcp.fullenrich.com, documented 7 May 2026, also live for Claude and ChatGPT. Search previews are free for the first 10 results; a full export costs 0.25 credit a record. FullEnrich’s waterfall model over 20-plus providers is a different bet than Lusha’s single-source database, and its MCP server inherits that same tradeoff.

Verdict

A team already on Lusha, comfortable with its credit system and mainly running lookups a rep could otherwise do through the browser extension, gets a real convenience upgrade from the MCP connector: the same data, without switching tabs, on whatever plan it already pays for.

Anyone picking a sales-intelligence vendor mainly for its AI-agent story should look past the marketing page’s own numbers first. Lusha’s MCP landing page can’t keep its own company count consistent within a single scroll, and its accuracy claims move depending on which of Lusha’s pages states them.

The gap sits in Lusha’s documentation: whichever page a reader lands on first states a different number, so the account settings are the one place to check directly.

A team that needs webhooks, Website Visitors matching or Decision Makers data inside the same AI workflow should stay on the REST API for those calls; MCP doesn’t cover them, and Lusha hasn’t said when or if it will.