An MCP server for B2B data is a service that exposes contact search and enrichment as tools an AI assistant calls on its own. Clu runs one at https://mcp.cleanlist.ai/v1/mcp, and a connected Claude session sees 14 tools, including search_people, search_companies, enrich_person, enrich_list, estimate_cost, create_list, add_leads_to_list and get_lists. Search costs 0 credits in Clu. A verified email costs 1 credit, a phone number 10, and both together 11. Nothing paid runs until estimate_cost returns a signed quote you can see first.
Last updated: August 10, 2026. Credit costs, plan prices, rate limits and the tool list on this page reflect Clu as of that date.
What is an MCP server for B2B data?
MCP is the Model Context Protocol, described at modelcontextprotocol.io as an open standard for connecting AI applications to external systems. An MCP server for B2B data wraps a contact search and enrichment engine in that standard: the Clu server wraps Clu people search at 0 credits and waterfall enrichment at 1 to 11 credits, so an assistant like Claude runs the query itself rather than you opening another tab. Clu's server is remote and hosted at https://mcp.cleanlist.ai/v1/mcp, so there is nothing to install locally and no API key to paste into a chat window. Claude talks to it over HTTP and authorizes with OAuth. The Clu server is in beta on Starter and above, and search calls cost 0 credits.
Which vendors actually ship an MCP server for B2B lead data?
Clu, Apollo, Clay, Lusha and Crustdata all publish first-party MCP servers, and ZoomInfo shipped native Claude and OpenAI Codex MCP integrations in its Q2 2026 8-K, filed August 5, 2026, under its ticker NASDAQ: GTM. Claiming one and serving one are different things: we probed the endpoints of 60 vendors and confirmed 39 responding servers in the GTM MCP server census. This is a contested surface with a public-company incumbent already on it. Clu differentiates on the money model instead: search costs 0 credits, estimate_cost returns a signed quote before any spend, email and phone enrichment is pay-for-results so a lookup that returns nothing is not charged, and the whole thing is self-serve from $49 a month with no annual contract. The longer ten-server side-by-side is in the Clu guide to sales and GTM MCP servers. For ZoomInfo's server on its own, read our ZoomInfo MCP review.
What tools does the Clu MCP server expose to Claude?
A connected Claude session shows 14 Clu tools in five groups. Search: search_people and search_companies, both 0 credits. Enrichment: enrich_person for one record, enrich_list for a whole list, plus add_person_enrichments and add_company_enrichments to attach enrichment to records you have already saved. Lists: create_list, add_leads_to_list, get_lists, list_leads_in_list, remove_leads_from_list and delete_list. Cost control: estimate_cost. Identity: whoami. Names map onto Clu REST API v2 endpoints, so search_people is POST /search/people and estimate_cost is POST /credits/estimate. The full Clu server publishes 30+ tools across roughly 30 v2 endpoints, and a given connector exposes a curated subset, so read tools/list in your own client before you write a prompt that assumes a tool exists.
How do I connect Claude to Clu lead data?
Open Claude, go to Settings, then Connectors, add a custom connector pointed at https://mcp.cleanlist.ai/v1/mcp, and complete the OAuth sign-in. There is no key to copy and no local process to run. It is a custom connector rather than a directory click because Clu was not listed in Anthropic's connector directory when we last checked on August 7, 2026, while comparable B2B data connectors were. The practical difference is two extra clicks, after which the OAuth flow, the tools and the credit wallet are identical. Revoking from the same Connectors screen cuts the agent off from both your Clu data and your credits immediately. Setup detail and the current beta gate live on the Clu MCP server page.
How do I connect Clu to Claude Code?
Run claude mcp add --transport http cleanlist https://mcp.cleanlist.ai/v1/mcp, then run /mcp inside Claude Code and complete the OAuth sign-in. Anthropic documents that syntax for remote HTTP servers at code.claude.com/docs/en/mcp and states to use /mcp to authenticate with remote servers that require OAuth 2.0 (fetched August 7, 2026). The Clu endpoint returns 401 until that grant exists, which is how Claude Code knows to prompt you. Decide scope up front: --scope local keeps the server on your machine, --scope project writes it to .mcp.json for everyone on the repo. The OAuth grant is per user, but the Clu credit wallet is shared across your whole organization, so a project-scoped server means teammates draw on the same balance.
What does each Clu MCP call cost in credits?
Clu charges MCP tool calls at the same rates as the app and the REST API. People search and company search cost 0 credits. A partial enrichment, meaning email plus LinkedIn plus title plus company, costs 1 credit. A phone-only enrichment costs 10 credits. A full enrichment, meaning email and phone together, costs 11 credits. A company enrichment costs 1 credit. Reading lists and checking your balance cost 0 credits. There is no separate MCP subscription and no per-seat fee for agent access. Credits come from one shared organization wallet, so an agent in Claude, a teammate working in the Clu app and a script hitting the API all draw from the same pool. Plan prices are on the Clu pricing page.
How does estimate_cost stop an AI agent from overspending?
Every paid Clu operation is gated behind a signed quote. The agent calls estimate_cost first, which returns the price of the job, your current balance, whether that balance covers it, and an HMAC-signed quote_id with a five-minute time to live. The Clu server rejects the paid call if the real cost would exceed the quote or the quote has expired, so an agent cannot silently start a job you never priced. Three things sit underneath that. Search and reads cost 0 credits, so an exploring agent burns nothing. Enrichment is pay-for-results, so a lookup that returns no email is not charged. Your balance is a hard ceiling, because the wallet is shared across the app, the API and MCP.
Can Claude search for leads in plain English through Clu?
Yes, by delegation. Claude reads your plain-English request and calls the structured Clu search_people tool with filters. The Clu server takes filters and does not take prose. Ask for VPs of Sales at US SaaS companies with 201 to 500 employees, and Claude maps that onto the title, seniority, country and headcount fields itself. This is the part worth understanding before you buy: the natural language lives in the client, and the data layer stays structured and auditable. Clu search defaults to 25 results per call, so an agent that wants volume has to ask for it explicitly, and every one of those searches costs 0 credits regardless of how many times the agent refines the filters.
Does the Clu REST API support natural language search?
No, and the distinction is load-bearing. Clu, our GTM agent in the app, and the Clu MCP server both accept plain English, because a model sits in front of the query and converts it. The public Clu REST API v2 takes structured filters only and has no natural-language search endpoint. If you are writing code against POST /search/people, you build a filter object, not a sentence. Programmatic REST access is gated to a Pro plan or above at $89 per seat a month, or an AppSumo Tier 5+ license, while MCP access starts on Starter. Anyone benchmarking vendors on "does the API understand English" should test the endpoint rather than the demo.
What can Claude filter on in Clu people search?
Clu People Search exposes 20+ distinct filter dimensions through search_people, and every one of them costs 0 credits to query. They cover name and LinkedIn URL, job title with contains, exact and similar match modes, seniority from Owner and CXO down to Entry, years in role, years of experience, country, state and city with include and exclude, skills, connections, certifications, recently changed jobs, degree type, school names, fields of study, past titles, past companies, company name, company HQ location, industry on the LinkedIn taxonomy, and headcount bands from 1 to 10 up to 10,001 and above. There are no intent, revenue, funding or technographic filters in Clu. The full inventory is on the Clu People Search page.
What can Claude filter on in Clu company search?
Clu Company Search accepts exactly 9 filters through search_companies, and it is worth knowing which three people assume and do not get. technologies and revenue_ranges are rejected by the Clu backend outright. funding_stage is accepted and carries data: series_b returned 10,498 companies on September 7, 2026. The trap is the enum casing, because Series B returns zero rows and no error. That silent-zero behaviour is the single most common way an agent workflow quietly returns nothing. Build company cohorts on the filters that carry data, then enrich, then score. If your ICP is genuinely defined by tech stack or funding round, Clu is the wrong server for that job and a vendor that sells that layer will serve you better.
How does Clu actually find the email once Claude asks for it?
Clu runs a waterfall rather than a single lookup. A request goes through 25+ providers in sequence, stopping at the first verified result, with dedicated email verification behind it. In the Clu 500-Lead Enrichment Benchmark 2026, run on 500 stratified B2B leads, the waterfall returned 98% verified email and 85% direct dial. Single-source providers returned 70 to 80% email and 30 to 60% phone on identical input. That gap is the reason the credit costs are shaped the way they are: 1 credit for an email that took several provider calls to confirm, and no charge at all when nothing verified comes back. More detail sits on the Clu waterfall enrichment page.
Can I import a CSV through the Clu MCP server?
No. CSV file import is portal-only, and none of the 14 tools on the Clu MCP connector as of August 10, 2026 accepts a file. Upload the CSV in the Clu importer at app.cleanlist.ai/lists, then come back to Claude and run enrich_list against the resulting list id with a signed quote from estimate_cost. That path is the one that scales, because bulk enrichment is asynchronous and one enrich_list call covers the whole list at 1 credit per verified email. Clu publishes no fixed per-call contact ceiling for MCP, so the job is bounded by your credit balance rather than a batch limit, and the agent polls until the run settles.
Can Claude sync leads to my CRM through the Clu MCP server?
No. None of the 14 tools on the Clu MCP connector as of August 10, 2026 syncs to a CRM, and neither AI research agents nor lookalike company search are exposed there either. Those jobs run in the Clu app. The honest shape of an agent workflow today is find, enrich and organize inside Claude, then push to HubSpot, Salesforce or Lemlist from the Clu portal or the REST API, where sync is priced at 0.2 credits per lead. If a vendor comparison tells you an agent takes a prompt all the way into your CRM, check its published tool list first. Clu would rather publish the boundary than have an agent fail halfway through a run.
If your CRM ships its own MCP server, connect it beside Clu in the same Claude chat: find and verify the contacts with Clu, then write them with the CRM's server. Our reviews cover the HubSpot, Salesforce, Pipedrive and Attio MCP servers, and the Gong MCP server adds read-only call context to the same chat.
What does a full Claude prospecting run look like end to end?
A realistic Clu run is four steps. Claude calls search_people with your filters, which costs 0 credits and returns 25 results by default. Claude calls create_list and add_leads_to_list to save the ones worth keeping. Claude calls estimate_cost for a partial enrichment on 40 saved leads, which quotes at most 40 credits and returns your balance plus a signed quote_id. Claude calls enrich_list with that quote and polls until the run settles. Pay-for-results means the actual charge can land below the quote when some records return nothing. The same 40 leads at full enrichment, email and phone together, would quote 440 credits, which is why the estimate step exists.
What are the rate limits on the Clu MCP server?
Clu enforces 60 requests per minute per organization, 30 requests per minute per API key, and a hard cap of 60 People Searches per UTC day per key. Those limits live on the REST API v2 endpoints, and MCP tool calls run against the same endpoints inside the same organization, so the per-organization ceiling and the daily search cap apply to agent traffic too. The 30-per-minute figure is scoped to a clapi_ API key rather than to an OAuth connector. Exceeding a limit returns a 429 with a Retry-After header. The daily search cap is the one that bites in agent workflows, because an agent looping over many narrow searches will burn 60 in an afternoon. Batch your filters into fewer, broader searches and page through the results. Clu publishes no uptime SLA and the MCP server is in beta.
Which Clu plans include the MCP server?
The Clu MCP server is in beta on self-serve accounts, and AppSumo licenses are excluded from the beta. MCP access starts on Starter, a lower gate than programmatic REST API access, which starts on Pro. Self-serve pricing is Free at $0 with 50 credits a month, Starter at $49 per seat a month with 750 credits per seat, and Pro at $89 per seat with 1,500 credits per seat. Enterprise is custom. Annual billing takes 25% off seat prices. There is no separate fee for agent access, so a Starter workspace can connect, run searches at 0 credits up to the 60-search daily cap, and spend a seat's 750 credits on 750 verified emails.
What can the Clu MCP server not do?
Clu publishes the boundary rather than letting an agent discover it mid-run. There is no CSV import over MCP, no CRM or sequencer sync over MCP, and no natural-language endpoint on the REST API. Clu REST API v2 has no webhooks, so asynchronous enrichment is polled rather than pushed. There is no intent data, no buying-signal feed, and no job-change alerting: "recently changed jobs" is a query-time filter, so an agent can ask who moved but nothing will notify you when someone in an existing list moves. There is no revenue, funding-stage or technographic filtering. Clu publishes no SOC 2 report and no uptime SLA, and there is no read-only agent key today.
Is my data safe when Claude calls Clu?
Whatever a Clu tool returns goes into the model's context, because that is how MCP works: the client hands tool results back to the model for its next turn. Verified emails and phone numbers Claude retrieves through Clu are handled under your agreement with the model provider, so a data-processing review has to cover the MCP client and Clu separately. The connector authorizes over OAuth scoped to your Clu organization, so no raw key sits inside a chat client, and revoking from Claude Settings, then Connectors, cuts access immediately. Clu publishes no SOC 2 report and no uptime SLA as of August 10, 2026, so wait for formal attestations rather than assuming them.
How do I run my first bulk enrichment through Claude?
Create a Clu account, which opens on Pro for 14 days with 250 credits and three seats, add https://mcp.cleanlist.ai/v1/mcp as a custom connector in Claude Settings, then Connectors, and ask Claude to find 25 people matching your ICP. That search costs 0 credits. Save the good ones to a list, then have Claude quote the enrichment with estimate_cost before it spends anything. For a list you already own, upload the CSV in the Clu importer at app.cleanlist.ai/lists first and run enrich_list against it. The 250 trial credits buy 250 verified emails, which is enough to judge the waterfall against whatever you are paying for today.
References & Sources
- [1]
- [2]
- [3]
- [4]
- [5]
Omni
Decagon
Distyl
WellSaid
Sanity
Distyl