MoltSets and our ongoing test of the $27/mo enrichment subscription

We ran the same 821 data and engineering profiles through MoltSets on September 28 and again on October 22, 2026. Usable emails, meaning an email on the person's own company domain, went from 356 to 411. That moves the usable rate from 43 percent to 50 percent. But coverage barely changed: 54 percent of profiles returned any email in September, 55 percent in October.

MoltSets is a headless contact-data API run by Adam Robinson of RB2B and Retention.com. Keys and billing live in an app. The data work runs through the API, an MCP server or a Chrome extension. It's built for terminals and agents. Paid plans start at $27 a month, and pricing follows rate limits instead of seats or credits. We pay for it. Our contact-enrichment waterfall today runs BlitzAPI first, then MoltSets, then Clay.

We're continuing this 821-profile test on a cadence while Adam Robinson and the MoltSets team build in public. Clay was our primary source earlier this year and now sits last in the waterfall, filling what the first two miss. The aim is to see the trajectory of service quality, and when MoltSets alone covers what MoltSets plus Clay cover together: 71 percent usable on this list in September. Then Clay drops out. The next run comes after November 1, 2026, the date the vendor gave for its 3-day data refresh.

What MoltSets is

MoltSets sells contact and company data. The app and console handle keys and billing, and the founder's October launch post says plainly that there's no UI for the data work. That runs through the API, the MCP server or the Chrome extension, called from a terminal, a script, an agent such as Claude Code or Cursor, or a Clay table.

The $27 plan, which the pricing page now calls solo, allows 1,000 enrichment records and 500 search records per five-hour window, capped at 5,000 enrichment and 2,500 search records a week, with 3 searches running at once. The rate-limits page also counts requests, at 5,000 enrichment and 2,500 search requests per five hours, and a miss spends no records though it still counts as a request. Carrier-verified mobile numbers cost extra, at $0.15 each, and the site still carries a public beta banner as of October 22, 2026. Our MoltSets tools page has the short version.

Why MoltSets prices by rate limit instead of seats

Apollo and ZoomInfo price mainly by seat, and the founder's launch post names seat pricing as the target. Clay prices by credit (we cover that in our credits-versus-actions piece). MoltSets prices by rate limit: how many records you can pull in five hours and how many searches can run at once. "We copied Claude," the founder wrote on September 25. A pricing article published October 12 describes credit tiers at $49 and $149 a month. But the vendor's pricing page on October 22 has no such plans. It lists four at $27, $97, $397 and $997 a month, with 1,000, 15,000, 150,000 and 225,000 enrichment records per five-hour window.

Seats count people with logins, and credits count lookups. But a rate limit caps throughput, which a script can saturate without anyone logging in. Run more searches at once than your plan allows (3 on $27) and you get a 429 with a 1-second retry. And hitting a five-hour window gets a 429 that says how long to wait. Mobile numbers left the base plans on October 2 and now cost $0.15 each, with 180 for $27.

The launch post names a buyer for each plan, from an individual rep working inside Claude at $27 to two named power users at $997, with a lead-gen agency running 10 accounts at $397. The October 16 post says marketing is paused except the newsletter and the occasional LinkedIn post. Our comparison of Apollo, ZoomInfo and Clay covers the seat-priced side.

The trajectory so far

• Around May 19: Adam Robinson announces MoltSets on LinkedIn as unlimited contact-data APIs for Clay and Claude Code. The date is decoded from the LinkedIn post ID.
• Late August: the beta starts, by our estimate from the newsletter archive.
• September 4: the newsletter says about 20 percent of deliverable emails didn't match the person's current role, and reports $265k in annual recurring revenue.
• September 18: a $397 plan arrives. The open problems named are the same 15 to 20 percent mismatch, no international data and no final email validation system. Revenue is reported at $360k.
• September 25: the newsletter defends rate-limit pricing against complaints that the limits are confusing.
• October 2: mobile numbers leave the base plans and drop to $0.15 each.
• October 8 to 9: the launch announcement says no UI and no waterfalls (the October 20 FAQ says the vendor's data waterfalls maximize first-call coverage), and promises that profiles refresh and emails re-validate every 3 days, live November 1.
• October 16: the waitlist is cleared and revenue is reported past $600k. The newsletter says profile data now refreshes weekly, not every 3 days. Richer profile fields and new filters ship, and a "rock-solid v1" is now expected January 1.
• October 20: the FAQ is updated. It says job changes are processed every 2 weeks and emails re-validate on a 90-day window.

Three refresh claims now sit side by side: every 3 days, promised for November 1; weekly, announced October 16; and every 2 weeks with 90-day email re-validation, in the October 20 FAQ. And the revenue figures are the founder's. The Baremetrics page he links to loaded only a sign-in screen, so we haven't verified them.

How we ran the test

The roster is 821 people at 10 US B2B technology companies, between about 170 and 3,000 employees each, carried over from our September 28 roster test: a consumer fintech, a payments infrastructure provider, a community messaging platform, a blockchain analytics firm, a mental-health care marketplace, an on-demand manufacturing marketplace, a restaurant payments platform, a digital nutrition-care provider, a freight logistics company and a salon software vendor. The people are the ones a data or AI vendor would sell to: 524 individual contributors in data, analytics and engineering (190 data scientists and 90 data engineers among them), 259 other senior executives at director level and above, and 38 data leaders. The companies and people are not named here; the roster was built for a client engagement. Each profile got one linkedin_to_business_email call on September 28 and one on October 22, 24 days apart. Both runs went to MoltSets alone.

An email is usable when it sits on the company's own domain or on a known alias, meaning a second domain the company itself uses for mail; three of the ten companies have one. An email on a different company's domain is stale, most likely the person's last role. Every returned email also carries a letter grade from A to F.

Clay wasn't re-run in October, and personal emails weren't tested. There was no bounce check. But "usable" tests the domain only, and the September ownership check covered Clay's guesses, not MoltSets' hits. And the sample is US-heavy (621 of the 821, by September's location read), with one provider and business email only.

The October run used 451 enrichment records against the plan's 1,000 per five hours. Median response time was 0.37 seconds, and 1.26 seconds at the 90th percentile.

Results: usable emails up from 356 to 411

Usable emails rose by 55, a 15 percent relative gain. But coverage of any email stayed flat, 447 to 451. The gain comes from 92 new usable finds and 30 emails that moved from another company's domain to the right one, less the 61 usable emails that disappeared and 6 that turned stale. And the stale share of what MoltSets found fell from 20 percent to 9.

Underneath, the 821 churned. The same email came back for 259 profiles (249 of them usable). A different email came back for 79, and another 113 profiles that missed in September hit in October. In the other direction, 109 hit in September and missed in October: 48 were stale emails MoltSets no longer returns, and 61 had been usable. Neither run found an email for 261 profiles.

Those 61 are the cost. Not a clean win. The test doesn't show whether the people changed jobs or MoltSets dropped the records.

The spread is wide. The payments infrastructure provider climbed from 53 usable of 107 profiles to 69, the mental-health care marketplace from 17 of 90 to 28, the blockchain analytics firm from 21 of 63 to 37. But the consumer fintech, the largest group at 217 profiles, slipped from 101 to 93. Five emails now come back as Grade F, such as a profile at the consumer fintech returned at a facebook.com address.

What the richer profile shows

On October 22 we also pulled 24 of the same profiles with include_experience set to true: 12 whose September email sat at another company's domain, 6 clean hits and 6 misses. Job history came back on 23 of the 24, with a median of about 11 roles. Connection count came back on 23, skills on 18 and a summary on 13. The company block carried a description and industry on 22 profiles, revenue on 16, and NAICS and SIC fields on all 24. The profiles also carried hashed business emails and a validated-at timestamp.

The profile's current company matched the company we hold for 22 of the 24, by name. One person now works at Nokia, and one profile wasn't found. All 12 profiles with a stale September email still show the right company. The problem was the email, not the profile.

Filters for NAICS code, LinkedIn industry and Canada with the state set to Ontario were accepted and returned results. But the weekly refresh may not have reached every record: last_updated_at on 3 search rows ran from 2022 to 2026. Small sample.

Where MoltSets sits in our waterfall

For contact enrichment our order today is BlitzAPI, then MoltSets, then Clay. BlitzAPI also supplies the rosters, job postings and department headcounts we work from, and this test says nothing about MoltSets on those jobs. Clay, our primary source earlier this year, now sits last and fills what the first two miss. Our outbound data stack piece and waterfall teardown cover the wider build, and the Clay page covers the tool.

In September we also ran the same 821 profiles through MoltSets first, then Clay on the misses. That reached 583 usable emails (71 percent), with Clay adding 227 people MoltSets couldn't cover. MoltSets alone reached 356 (43 percent) that day. But Clay wasn't re-run in October, so 71 percent is the September mark for the pair.

Before a Grade A email goes into a sequence:

Coverage has two caveats. Personal email is US-only (September 18 newsletter, repeated in the October 20 FAQ), while the FAQ says the rest of the data isn't. Our sample is 621 US profiles of 821, so it says nothing about coverage abroad. But the 3-day refresh is still a promise.

Our view and the next run

In our opinion, a headless data vendor with a price set by rate limit may pressure seat-priced incumbents, because the person configuring it is an engineer or an agent, and neither needs a seat. But whether the pressure lasts depends on the data. The vendor admitted a 15 to 20 percent stale-role problem in September. Our October run shows 9 percent of found emails at another company's domain, down from 20. That's movement in the direction promised. But it also cost us 61 usable emails.

For a buyer today, we'd call $27 a month worth it as a second source behind whatever runs first, and not yet as the only source. But half of our list came back usable, and the other half needs another source. MoltSets alone reached 50 percent in October, against 71 percent for MoltSets plus Clay in September.

Three things would let Clay drop out of our waterfall: MoltSets alone at or above the 71 percent that MoltSets plus Clay reached on the same 821 in September, the 3-day refresh live, and a published bounce rate. And international coverage would matter too. The September 4 newsletter aimed for it by the start of Q4, and the October 20 FAQ says coverage isn't US-only, a claim our US-heavy sample leaves untested.

We're re-running the same 821 profiles on a cadence. The next run is after November 1, 2026, the first date the 3-day refresh can be live. Each run will be added to this page with its date and numbers.