Lusha and Apollo Are Moving Data Enrichment Into the AI Chat Window

Key takeaways

• Lusha and Apollo.io both run official MCP servers, so prospect search, enrichment and (in Apollo's case) outreach can happen inside a Claude conversation.
• In our query of Apollo's database, Indonesia has the largest pool of manager-and-above records in Southeast Asia, yet only 24.6% carry a verified email status.
• Lusha charges 5 credits for a phone number and 1 for an email. Apollo's people search is credit-free. That asymmetry should shape how you design an AI enrichment workflow.
• MCP does not add a new credit pool. It makes spending existing credits much faster, which turns guardrails into a design requirement.

Enrichment used to be a tab you switched to. Now it is a tool the model calls

For most of the last decade, data enrichment was a separate ritual. A marketer exported a list from a CRM, uploaded it to an enrichment vendor, waited, downloaded the result, cleaned the columns and pushed it back. A sales rep did the same thing one contact at a time through a browser extension sitting on top of LinkedIn.

The Model Context Protocol, an open standard that lets AI assistants call external tools, collapses that loop. It is the same shift we described in how agentic AI is transforming marketing workflows, now applied to contact data. When Apollo.io launched its connector in Claude's directory on 24 February 2026, CEO Matt Curl framed the problem directly: "AI conversations are where work increasingly begins, but revenue teams experience friction in switching systems when it comes to outbound execution." Lusha had published its own MCP server earlier, in July 2025, and has expanded it since.

The practical result is that a request like "find twenty heads of marketing at Jakarta fintech companies with 50 to 200 employees and get me their work emails" is no longer a brief for a junior analyst. It is a prompt that triggers a chain of API calls: filter, search, match, reveal, write back.

290M+ verified B2B contacts claimed by Lusha, across 29M+ companies

240M contacts claimed by Apollo.io, across 30M companies

50+ actions exposed by Apollo's MCP server, from search to sequence enrolment

Two databases, two different philosophies

It is tempting to treat Lusha and Apollo as interchangeable contact databases. Their product design says otherwise.

Lusha is built as an enrichment layer. Its centre of gravity is the reveal: you already know who you want to reach, and Lusha gives you a verified email, a direct dial and firmographic context. It has added prospecting filters, intent signals, lookalike search and scored recommendations, but the product still reads as a precision instrument that plugs into whatever CRM and outreach stack you already run. Lusha reports roughly 90% accuracy on contact details, and independent testers have singled out its direct-dial phone data as a relative strength.

Apollo.io is built as an operating system for outbound. The database is one module; sequences, tasks, dialler, call recordings, deal tracking and analytics sit alongside it. Its MCP server reflects that breadth. Beyond search and enrichment, it can create contacts and accounts, enrol people into sequences, create tasks, pull conversation transcripts and report on campaign performance.

The difference matters in an AI context. An assistant connected to Lusha is a very good researcher. An assistant connected to Apollo can also be an operator, which is more powerful and, for the same reason, needs tighter permissions.

Indonesia has the biggest pool in the region, and the thinnest verification

Global database sizes say little about the market you actually sell into. So we queried Apollo's people database directly through its MCP server on 29 September 2026, counting records located in each of six Southeast Asian countries at manager seniority and above, then counting how many of those carry a verified email status. People search in Apollo does not consume credits, which makes this kind of coverage audit free to repeat.

Two things stand out. First, Indonesia's raw pool of roughly 639,000 senior records is about 55% larger than Singapore's. Second, fewer than one in four of those Indonesian records carries a verified email, against almost one in two in Singapore. Vietnam is thinner still.

The likely explanations are structural rather than a vendor failure: a higher share of Indonesian professionals use personal Gmail or Yahoo addresses for work, company domains change more often at smaller firms, and many mid-market businesses run on WhatsApp rather than email for first contact. Whatever the cause, the implication for teams selling into Indonesia is concrete. A list that looks large on screen shrinks sharply once you filter for deliverability, and phone enrichment carries more weight here than in markets where cold email still does most of the work.

The number that matters is not how many records a vendor has in your market. It is how many of them you can actually reach.

Data decays, but slower than the industry likes to claim

Enrichment is not a one-off purchase because the underlying data moves. The figure most often repeated in vendor marketing is that B2B data decays by roughly 30% a year. Lusha tested that claim against its own records in August and September 2026, measuring only job changes, and found a slower but still meaningful rate: about 1% a month in the United States, with large differences between countries.

Lusha notes that the popular 30% figure appears to conflate a two-year rate (25.67% in its US sample) with an annual one. It also concedes that job changes are only one failure mode; when title changes, number porting and email migrations are counted, the real rate is higher. HubSpot's long-cited benchmark of 22.5% annual decay sits between the two.

For planning purposes, the useful rule is simple. A contact list enriched once is noticeably stale within a year, and the decay is faster in markets with high job mobility. The UK rate is roughly 1.75 times the US rate. There is no published equivalent for Indonesia yet, which is itself a gap worth noting for anyone building a regional data product.

How Lusha works, step by step

Lusha offers three entry points: a web app, a Chrome extension that overlays LinkedIn and company websites, and an API. The MCP server sits on top of the same API.

Using the Lusha platform

1. Create an account with a work email at lusha.com. The free plan gives 40 credits a month, enough to test data quality on a sample of your own target accounts.
2. Install the Chrome extension. When you open a LinkedIn profile or a company website, the extension shows whether an email or phone is available before you spend anything.
3. Reveal selectively. A verified email costs 1 credit; a phone number costs 5. Decide per contact whether a phone is necessary.
4. Prospect in bulk from the Prospecting tab: filter by location, seniority, department, company size, industry and technology. Contact and company searches cost 1 credit per 25 results returned.
5. Enrich an existing list by uploading a CSV (name plus company or domain, or LinkedIn URL). Lusha matches each row and appends the fields you choose.
6. Connect your CRM. Lusha integrates with HubSpot, Salesforce and other systems so revealed contacts sync without an export step.

Connecting Lusha to Claude via MCP

1. Confirm your plan. Lusha says MCP is available on Sales Starter and above, and it draws from the same credit pool as the API.
2. In Claude (web or desktop), open Settings → Connectors, find Lusha in the directory and choose Add. Complete the OAuth sign-in with your Lusha account.
3. Start a new conversation and check that Lusha appears in the tools menu.
4. Ask Claude to check your balance first. The account_usage tool is free and returns remaining credits, rate limits and the per-action price list.

For Claude Code or other MCP clients that do not use OAuth, generate an API key in the Lusha dashboard (API → Manage API Keys) and pass it as a header:

claude mcp add --transport http lusha https://mcp.lusha.com \
  --header "x-api-key: YOUR_LUSHA_API_KEY"

Lusha's documentation lists 23 tools across contacts, companies, prospecting, signals and account management. The live server we connected also exposed newer tools for working tables, buying-group search and exporting results straight into a CRM.

How Apollo works, step by step

Using the Apollo platform

1. Sign up with a work email. This matters more than it sounds: Apollo's documentation states that free personal accounts cannot use people or company search and enrichment through MCP.
2. Search the database under Search → People or Companies. Filter by title, seniority, location, headcount, headcount growth, funding, technologies and active job postings. Searching does not spend credits.
3. Save to a list and enrich. Revealing an email typically costs 1 credit; mobile numbers draw from a separate, smaller mobile-credit allowance.
4. Enrich existing data by importing a CSV or connecting your CRM, then running bulk enrichment on the records that are missing fields.
5. Build a sequence that mixes emails, calls, LinkedIn steps and tasks, and enrol the enriched contacts.
6. Review analytics for reply, bounce and meeting rates, and feed that back into your targeting filters.

Connecting Apollo to Claude via MCP

1. In Claude, go to Settings → Connectors, select Apollo and authenticate with OAuth. No API key is needed, and access follows your Apollo user permissions and plan limits.
2. For Claude Code, run the command below, then type /mcp and complete the OAuth flow in the browser.

claude mcp add --transport http apollo https://mcp.apollo.io/mcp

Apollo also publishes a plugin for Claude Code and Cowork on GitHub (apolloio/apollo-mcp-plugin). Headless setups, such as a scheduled job, can authenticate with a master API key in the X-Api-Key header, but that key grants workspace-level access and should be treated like a production secret.

What an enrichment workflow looks like inside Claude

The interesting part is not either connector on its own. It is what happens when the model can use both in one conversation and choose which call to make based on cost. Here is a pattern that reflects the credit economics described above.

1. Build the list where search is free

"Using Apollo, find marketing directors and heads of growth at companies headquartered in Jakarta with 51 to 500 employees in fintech or e-commerce. Show me the count first, then the top 30 by company headcount growth over the last six months."

Apollo's people search returns names, titles and employers without spending credits. Asking for the count first lets you tighten filters before anything is revealed.

2. Enrich emails in bulk, with a budget

"Enrich emails for these 30 people in Apollo. Before you start, tell me the estimated credit cost and stop if it exceeds 40 credits. Skip anyone whose email status is not verified."

Apollo's MCP responses include credit estimates and actual spend, so an explicit budget in the prompt gives the model a clear stopping rule.

3. Use Lusha only where a phone number changes the outcome

"For the 8 contacts with no verified email, check availability in Lusha first. Only reveal phone numbers for directors and above, and show me the credit total before revealing."

At 5 credits per phone number, reserving Lusha for the gaps is usually cheaper than running the whole list through it. In a market like Indonesia, where verified email coverage is low, this step often recovers the most valuable contacts.

4. Write back and act

"Create these contacts in Apollo, tag them 'Jakarta fintech Q4', and add the verified-email contacts to the draft sequence called 'Regional growth intro'. Do not activate the sequence."

This is where Apollo's broader action set comes in. Keeping the final activation step manual is a sensible default while a team builds trust in the workflow.

The credit math, made explicit

Take 200 senior contacts that need both an email and a phone number. On Lusha alone, the search costs 8 credits (200 ÷ 25), the emails 200 credits and the phones 1,000 credits: 1,208 credits in total, with phones accounting for 83% of the spend. Moving the search to Apollo and revealing phones only for the 25% of contacts that truly need a call cuts the Lusha bill to about 250 credits. The saving comes almost entirely from being selective about phone numbers.

What changes when an AI agent holds the credit card

Neither vendor charges extra for MCP access. Lusha is explicit that MCP "uses the same credits as the regular Lusha API." What changes is velocity. A person clicking reveal in a browser extension spends credits at human speed. A model working through a list can spend a month's allowance in a single turn if the prompt is vague.

Three guardrails are worth building in from the start:

• Count before you reveal. Ask for totals and cost estimates, then approve. Both servers support checking balances without spending.
• Separate read and write permissions. Apollo's MCP can create records and enrol contacts in sequences. Give that scope only to people who already own outbound.
• Log what was enriched and why. Indonesia's Personal Data Protection Law (UU PDP, Law No. 27 of 2022) applies to personal data, including work contact details tied to an individual. Keep a record of the lawful basis for processing, and honour opt-out requests quickly.

Where this leaves marketing and sales teams in Indonesia

The shift from dashboards to conversations lowers the skill barrier for list building. A founder or marketer who has never written a Boolean search string can now describe an ideal customer in plain language and get a filtered, enriched list back in minutes. That is a genuine productivity gain, especially for small teams without a dedicated RevOps function, and it rewards teams that have already done the work of segmenting their market properly.

It also raises the importance of judgement. When enrichment becomes cheap to request, the constraint moves to knowing which contacts are worth reaching, which channel fits the market, and when the data is too thin to trust. Our Apollo query suggests that in Indonesia, the second and third questions deserve more attention than they get: the pool is large, but reachable, verified contacts are a much smaller slice of it.

Frequently asked questions

What is data enrichment in B2B marketing?

It is the process of filling in a thin record, such as a name and company domain, with verified attributes: work email, direct phone, job title, seniority, company size, industry, technology stack and buying signals. Lusha and Apollo do this by matching your record against their own databases.

Is Lusha or Apollo better for data enrichment?

They are optimised for different jobs. Lusha is a focused enrichment layer with a strong reputation for direct phone numbers. Apollo combines a large database with sequencing, tasks and analytics. Many teams search in Apollo and enrich selectively with Lusha.

How do I connect Lusha or Apollo to Claude?

Both are in Claude's connector directory. Open Settings → Connectors, add the tool and complete the OAuth sign-in. In Claude Code, use claude mcp add --transport http with https://mcp.apollo.io/mcp or https://mcp.lusha.com.

Do MCP calls cost extra credits?

No separate pool is used. Both vendors charge MCP calls against your existing plan credits. The difference is that an AI agent can spend them much faster, so set budgets in your prompts.

How accurate is B2B data for Indonesia?

Coverage is broad but verification is thinner. In our 29 September 2026 query of Apollo, Indonesia had about 639,000 manager-and-above records, but only 24.6% carried a verified email status, compared with 49.3% in Singapore.

Is it legal to use enriched contact data in Indonesia?

UU PDP (Law No. 27 of 2022) governs personal data, including work contact details tied to a person. Document a lawful basis, honour opt-outs and keep enrichment scoped to legitimate business contact. This is general information, not legal advice.

Can I use Lusha and Apollo together?

Yes. A cost-efficient pattern is Apollo's free search to build the list, Apollo enrichment for emails, and Lusha for the smaller set of contacts that need a direct phone number. Inside Claude, both connectors can run in the same conversation.

Keep reading

• How Agentic AI Is Transforming Marketing Workflows
• How Autonomous Systems Are Changing the Way Brands Grow
• What Market Segmentation Means, with Types, Data, and Examples
• What Is Personalization? The Data Behind It and How to Apply It
• Performance Marketing Is Results-Based Advertising
• How Digital Brands Can Acquire and Retain Customers More Efficiently