# Explorium Ends Dec 2. Export the Prompts, Not Just the Contact Lists

_Explorium shuts AgentSource and Vibe Prospecting Dec 2. What to export beyond the CSVs, and how to replace account-plus-person search without two databases._

- Canonical URL: https://www.refolk.ai/blog/explorium-shutting-down-export-prompts-alternatives
- Published: 2026-11-14
- Reading time: 10 min
- Keywords: explorium alternatives, vibe prospecting alternative, explorium shutting down, agentsource alternative, B2B prospecting MCP server

Explorium's founders announced the week of Nov 3, 2026 that the company is closing. AgentSource and Vibe Prospecting stop working on Dec 2. If you used either one, you were really asking one question: which accounts match my criteria, and who at those accounts should I talk to? Most of the replacements being recommended now answer only half of it. Before you export a CSV and sign up for the first tool on a roundup, work out which half you're about to lose.

## What exactly is shutting down, and when?

AgentSource, Vibe Prospecting and all data exports end on December 2, 2026. Explorium will delete all user data and account information no later than December 31, 2026. After Dec 2 you can't pull anything out, and after Dec 31 there's nothing left to pull.

The announcement date depends on which source you read. PredictLeads dates it to November 3. CTech and CEO Medium say November 4, when co-founders Omer Har and Or Tamir posted their farewell. Either way, users got about four weeks.

The dates that matter:

- **Dec 2, 2026:** AgentSource, Vibe Prospecting and data exports stop.
- **Within 30 days:** unused credits are refunded, according to the ReachKit directory listing.
- **By Dec 31, 2026:** all user data and account information is deleted.

This isn't a niche tool going dark. The founders say Vibe Prospecting grew out of "the first B2B data enrichment MCP" and passed 350,000 registered users. Explorium raised $127 million from investors including Insight Partners, Zeev Ventures, Emerge and F2 Capital. Geektime reports that more than 70 employees are expected to lose their jobs.

> **350,000+** Registered Vibe Prospecting users, per the founders
>
> The founders' figure for the in-chat prospecting app that grew out of Explorium's MCP.

## Why do most Explorium alternatives cover only half the job?

Explorium sold company discovery and contact finding in one product, and most replacements specialize in one or the other. If you swap in a company-signal feed or a contact database, you get half your workflow back. The other half becomes a second tool to buy and a join you now have to maintain yourself.

Here is what Explorium marketed: company discovery across 150M+ profiles, contacts across 800M+ professionals, and 18 buying-signal categories. The pitch said outright that you could "skip stitching 2-3 tools together." That one query was the product.

Now look at what the Explorium alternatives roundups actually recommend. One of the most visible comes from PredictLeads, and it recommends PredictLeads "for teams replacing Explorium's company intelligence and signal use cases": firmographics, hiring, technographics, news, financing, connections and website evolution. It lists no contact data. Explorium's own comparison pages had already called PredictLeads "pure-play company intelligence data infrastructure." A vendor-written list will always put the vendor first, so check whether each recommended tool returns *people*, not just whether it appears on the list.

Side by side:

| Option | Returns accounts | Returns people | Catch |
|---|---|---|---|
| Explorium (until Dec 2) | Yes, 150M+ profiles | Yes, 800M+ professionals | Shutting down |
| PredictLeads | Yes, signals and firmographics | No | Company side only |
| Apollo MCP | Contact-first | Yes, 230M+ contacts | Third-party wrappers exist |
| ZoomInfo MCP | Yes | Yes | Read-only; tool count disputed |
| Clay | Enrichment layer | Via waterfall | Sits on top of a database |

Apollo launched its MCP server in February 2026, which gives Claude and ChatGPT access to its 230M+ contacts. ZoomInfo's server covers both sides. One review lists six tools (find, enrich and research for accounts and for contacts), another counts 23, and the server is read-only.

The default advice says the quiet part out loud. One 2026 guide calls Apollo or ZoomInfo as the contact foundation "plus Clay on top for custom enrichment and signal workflows" the dominant pattern for serious B2B teams. That's two data sources and a glue layer. It's the stitching Explorium promised you could skip.

## What does running two databases actually cost?

Running a company feed and a contact database side by side costs you a person, because someone has to keep the join between them working. In Refolk's index of professional profiles, that kind of person is scarce: 647 people in the US hold a GTM Engineer title, against 45,736 who hold an SDR or BDR title.

A GTM Engineer (go-to-market engineer) builds and maintains the data plumbing behind outbound: enrichment workflows, routing rules, and the code that connects a company list to a contact list. When Explorium answered "account, then person" in one call, far fewer teams needed that job. Without Explorium, somebody has to own these tasks:

- **Domain matching.** Subsidiaries, rebrands and regional domains mean the company feed and the contact database often disagree about what counts as the same account.
- **Title mapping.** "Head of RevOps" in one source and "Revenue Operations Lead" in another have to be treated as one target.
- **Re-enrichment.** People change jobs, so contacts attached to an account go stale and need refreshing on a schedule.

> **~71** US SDRs and BDRs for every US GTM Engineer
>
> Derived from Refolk's index: 45,736 SDR/BDR titles against 647 GTM Engineer titles.

That ratio explains why stitched stacks fail without anyone noticing. Most prospecting teams don't have a person who owns the join. A domain mismatch quietly drops accounts, the list still looks full, and the SDRs keep sending. The damage shows up later as a thinner pipeline, and by then nobody connects it to the data.

| Segment (Refolk's index, Nov 2026) | Count | Note |
|---|---|---|
| GTM Engineer, US | 647 | Index count |
| GTM Engineer, UK | 97 | Index count |
| US:UK GTM Engineer ratio | 6.7x | Derived |
| SDR/BDR, US | 45,736 | Index count |
| US SDRs/BDRs per US GTM Engineer | ~71 | Derived |
| UK GTM Engineer sample at Clay | 20% (5/25) | Small sample, directional only |

Outside the US, the pool is smaller still. The UK has 97 GTM Engineers in Refolk's index, and 14 of a 25-profile sample are in London. In that sample, 5 work at Clay and 2 at Instantly.ai. Current employers in the US sample include Clay and Cognition. The sample is small, but it points one way: a real share of the people who build this glue work for the companies that sell it.

If your answer to Explorium's shutdown is "hire someone to own the Apollo-plus-Clay pipeline," expect a long search for a hard-to-fill role. You'll be recruiting from a small pool, and part of it is already employed by the vendors you're paying.

## What should you export from Explorium before Dec 2?

Export the logic behind your lists first and the rows last. The rows start going stale as soon as you download them, because people change jobs. The filters, signals and prompts that produced the rows are the part you can't rebuild from memory, and they work in any tool you move to.

Work through your account in this order:

1. **Saved queries and filter definitions.** For each saved search, copy every condition: industries, headcount bands, geographies, technologies, exclusions. Screenshots are fine if there's no text export.
2. **Signal choices.** Explorium advertised 18 buying-signal categories and 80+ signal types. Write down which ones you used for each list and why, for example "hiring for engineering" or "recent financing."
3. **Agent prompts.** Copy every prompt you typed into Vibe Prospecting or wired into AgentSource, word for word. These are your ideal-customer definitions in their most usable form.
4. **What converted.** Note which lists or prompts booked meetings and which didn't. This is the institutional knowledge that leaves with the tool.
5. **The CSVs.** Keep them as suppression lists (people you've already contacted) and as a baseline to compare against your replacement.
6. **Your credit balance.** Record it so you can check that the refund arrives within 30 days.

Next, rewrite each saved query as one plain-English sentence. A filter stack like this:

- Industry = fintech
- Employees = 200 to 1,000
- Region = UK
- Signal = hiring (engineering)
- Title contains "RevOps"

becomes: *Heads of revenue operations at UK fintech companies with 200 to 1,000 employees that are hiring engineers.*

That sentence doesn't depend on any vendor. You can paste it into any AgentSource alternative or any chat-based tool, or hand it to a contractor. A filter stack tied to one product's field names can't move like that.

> The CSV is a snapshot. The sentence that produced it is the asset, and you can rerun it anywhere.

## How do you rebuild the account-to-person query without two databases?

Put the account conditions and the person conditions into a single request, and use a tool that resolves both together. If you split it into "get the companies from tool A, then find the people in tool B," you've rebuilt the join and taken on the maintenance work from the previous section.

Geektime's diagnosis of why Explorium failed supports this: as AI tools spread, the capabilities that had set it apart became available in free chatbots. If a chatbot could copy the interface, the interface was never the moat. The lasting asset is your definition of who you're looking for, written clearly enough that any capable system can act on it. That's why the prompts matter more than the rows.

A good account-to-person sentence has four parts:

- **The person first.** Role, seniority, function. "Heads of revenue operations," not "companies with RevOps teams."
- **The account as conditions on that person's employer.** Industry, size, region.
- **The signal as a fact about the account.** "That are hiring engineers," "that raised a round this year."
- **Exclusions.** Existing customers, and anyone on the suppression list you saved from your old CSVs.

[Refolk](/) is built to take that kind of sentence as is. You describe the person and the account together and get back a ranked shortlist, without matching domains between two exports.

> **Example search:** "Find heads of revenue operations at UK fintech companies with 200 to 1,000 employees that are hiring engineers right now." - [run it on Refolk](https://www.refolk.ai/s/cqt7w4ze1t)
>
> A ranked shortlist of named people at accounts that fit, with each profile and the reason it matched.

Test any Vibe Prospecting alternative the same way. Take three of your best-converting Explorium prompts, run them word for word in the new tool, and compare the results with the CSVs you exported. If the tool needs you to break the prompt into two steps, it covers one side. If the results come back with people at the wrong accounts, it covers both sides badly. I'd hold Refolk to the same test before trusting it with a live sequence.

## How should you vet a B2B prospecting MCP server before moving to it?

Before you move a workflow onto any B2B prospecting MCP server, check who maintains it, what it returns and what happens to your data if it shuts down. Explorium shows that even the original, first-party server can disappear on about four weeks' notice.

A ZoomInfo-published guide warns that several Apollo MCP servers on the market are third-party wrappers rather than Apollo's own, and that they "vary in maintenance and authentication." A wrapper that stops being maintained fails the same way a stitched stack does: quietly.

Before you switch, check:

- **Who runs it.** Apollo's own endpoint is mcp.apollo.io/mcp. Anything else is somebody else's code sitting between you and the data.
- **What it returns.** Accounts, people, or both in one call.
- **Read or write.** ZoomInfo's server is read-only. Apollo's also handles sequences.
- **Portability.** Whether your queries and prompts can be exported as plain text.
- **Shutdown terms.** How long you'd have to export, and when data is deleted. Explorium gave about four weeks to export and set Dec 31 as the deletion deadline.

Explorium did one useful thing for its users: it proved that a natural-language request returning both the account and the person is a workflow people want. 350,000 of them signed up for it. Keep that workflow. Save the sentences that defined it, and choose a replacement that can answer them in one step.

## FAQ

### When does Explorium stop working?

AgentSource, Vibe Prospecting and all data exports end on December 2, 2026. Explorium will delete all user data and account information no later than December 31, 2026. The shutdown was announced the week of November 3, 2026 (PredictLeads gives Nov 3; CTech and CEO Medium report Nov 4). Anything you haven't exported by Dec 2 should be treated as lost.

### What is the best Vibe Prospecting alternative?

Pick one that returns both the account and the person from one request. Otherwise you're replacing one tool with two. PredictLeads covers company signals but not contacts. Apollo's MCP is contact-first. ZoomInfo's MCP covers both but is read-only. Test any candidate by running your best old Explorium prompts word for word and comparing the output with your exported CSVs.

### Do I get my unused Explorium credits back?

Yes. According to the ReachKit directory listing, the platform runs until December 2, 2026, and unused credits are refunded within 30 days. Write down your credit balance before Dec 2 so you can check the refund amount when it arrives.

### Is an MCP server enough to replace AgentSource?

Only if it answers the full query. An AgentSource alternative that only finds accounts, or only finds contacts, leaves you building the join yourself. In Refolk's index there are about 71 US SDRs and BDRs for every GTM Engineer, so most teams don't have anyone to maintain that join. Also check that the server is first-party, since third-party wrappers of popular data sources vary in maintenance and authentication.

---

_From the Refolk blog. More posts: https://www.refolk.ai/blog/index.md_