Gong vs Clari: how long it takes to get your own calls back out

On December 3, 2025, Clari and Salesloft announced their merger had closed and named Steve Cox chief executive of the combined company. That was a change of plan. The August 7, 2025 announcement had said Andy Byrne, Clari’s co-founder, would lead it. We pulled the seven pages ranking for gong vs clari on 20 August 2026 and read each one; not one of them mentions that the merger closed, and the freshest carries an update stamp thirteen days old.

So that’s a reason to write another gong vs clari post. It isn’t the interesting part.

The interesting part is that both vendors publish enough first-party detail to answer a question none of those seven pages asks: how long would it take you to get your own calls, transcripts and forecast history back out, and does that fit inside the window before the data is destroyed? You can work most of it out with a calculator. We did, on 20 August 2026, and the numbers are below along with the parsing we ran.

Disclosure first, because six of those seven pages are published by companies selling something adjacent and don’t say so. Craqly is a real-time meeting assistant. We compete for some of the same budget. Re-run the arithmetic yourself. The inputs all link back to the vendor page they came from.

Gong vs Clari: what each one lets you read before you sign

Gong’s public API introduction is prose. It tells you the rate limits and the auth scheme, and then, on the companion overview page, it says to see the API documentation and links out to gong.app.gong.io/settings/api/documentation. We fetched that URL on 20 August 2026. It returns 9,033 bytes of application shell titled “Gong public API docs” with no endpoint content in the HTML at all, because the reference renders inside the product for signed-in customers.

Clari’s two APIs go the other way. The Copilot spec sits at a raw URL you can curl with no account: 52,659 bytes of OpenAPI 3.0.1. The Revenue API spec is embedded inline in the developer portal page, 129,194 bytes of it, version 5.0.0 on OpenAPI 3.0.0. We parsed both with a nine-line regex script (paths matched on ^ (/[^\s:]+):, operations on the verb keys under them) and counted.

Thirty-three documented operations you can read before you buy, against zero. I’d rather have the gated reference and a good product than an open spec and a bad one, so treat this as a due-diligence signal rather than a scoreboard. But it does decide who can price the exit before signing, and that turns out to matter.

One finding fell out of the parse that we weren’t looking for. Clari’s Revenue API contains no call endpoint and no transcript endpoint. The word “transcript” appears zero times in the whole 129KB spec. Conversation data lives in Copilot, behind a different base URL, a different auth header pair, and a separate rate-limit budget. If you buy the merged platform expecting one API over your revenue data, that seam is visible in the specs today.

How long does it take to export your calls?

For Clari Copilot, you can compute it exactly from the public spec: one listing request per 100 calls plus one request per transcript, against a documented ceiling of 100,000 requests a week. A 25,000-call archive costs 25,250 requests, roughly 42 minutes at the documented 10 per second. For Gong you cannot compute it from public documentation, and in a gong vs clari evaluation that asymmetry is itself the finding.

The Copilot spec is blunt about the shape. Its /calls endpoint description says you can fetch only 100 calls in page, and that to get transcript, deal live stage you fetch each call individually using the call-details endpoint. So the cost of draining N calls is ceil(N/100) + N requests. Here’s what that looks like against the documented weekly quota.

Two things jump out. The per-second limit almost never binds. Ten a second burns a full week’s 100,000 in under three hours of wall clock, so the weekly quota is the real governor, and a mid-size archive of 100,000 calls tips you over it by about a thousand requests. That’s an annoying place to land. You finish 99% of the export, then wait until Sunday 0 GMT for the counter to roll.

The thirty-day clock on the Gong side

Gong’s data retention policy is the clearest statement of switching cost either vendor publishes. Standard retention is the lesser of three years and the period you are a customer. And then: if you cease being a Gong customer, your data will be irreversibly deleted in 30 days. The page also warns that if a retention period is shortened and calls are purged, Gong will not be able to restore those calls.

Put that next to the documented 10,000 API calls per day and you get a hard ceiling. Thirty days at 10,000 a day is 300,000 requests, total, for the whole company, to retrieve everything you want to keep. If Gong’s transcript retrieval costs roughly what Clari’s does, one request per call plus paging, 300,000 requests covers something near 297,000 calls. If it costs two requests per call, you’re closer to 148,000. I don’t know which, and I can’t check, because the endpoint reference is the thing behind the login. That’s a genuinely unresolved cell and I’d rather flag it than guess a number the way two of the seven ranking pages guessed a seat price.

Both limits are documented defaults, not walls. Gong’s own page says to contact support if you need them changed, and Clari’s export quota description says the monthly quota is at the organization level and can be adjusted as needed, and to work with your CSM. So the honest framing is: this is what you get without negotiating, on the day you are least able to negotiate, which is the day after you’ve told them you’re leaving.

Ask for the raised limit in writing, in the contract, before you sign. Not in the offboarding email.

Where a small team lands

Neither of these platforms is built for six people. Gong self-reported on May 12, 2026 that annual recurring revenue passed $500 million, growing over 55% year over year, with more than 5,000 customers and half of the Fortune 10. Those are unaudited numbers from a private company, so take them as direction rather than fact. The direction is enterprise.

If what you actually need is help during the call rather than an archive to mine afterwards, the archive question stops being the deciding one. Craqly is priced for that shape of team: Starter is free with 20 credits a month, Pro is $38/mo or $19/mo billed yearly with 180 credits, Pro Max is $59/mo or $39/mo yearly with 420, and Ultra is $99/mo or $69/mo yearly with 780. One credit is one minute of live session, and credits reset every month. Yearly saves 20%. Stealth mode is on every plan including the free one, which is the part people assume is paid-only and isn’t. There’s an Enterprise tier above that if you need it. If you want the longer version of that trade-off, we wrote it up in Gong alternatives for small sales teams, and there’s a direct Craqly and Gong comparison too.

Pricing: neither one will tell you

Pricing is where a gong vs clari comparison earns or loses its reader. We fetched both pricing pages on 20 August 2026 and counted dollar signs. gong.io/pricing contains zero dollar figures. It’s a quote form that says licenses are priced per user, that there is a platform fee based on the number of users supported, and that you can integrate your existing tech stack for free. The team-size dropdown offers four bands: 1 – 50, 51 – 1,000, 1,001 – 9,999, and 10,000+. No trial is listed.

clari.com/pricing also contains zero dollar figures. Its headline is “Priced to run revenue, not hurt the bottom line”, and its one substantive pricing claim is an out-of-the-box solution with no extra platform fees for integrations or continuous support. Read against Gong’s stated platform fee, that line is doing deliberate work.

Two of the seven ranking pages print per-seat figures anyway. The two disagree with each other by more than double, one of them labels its own table an estimate, and neither number appears anywhere on the vendor page it purports to describe. Numbers on a comparison page should carry the URL they came from or they shouldn’t be there.

Same caution applies to accuracy claims. Clari’s forecast product page says Clari Forecast helps teams achieve 98% forecast accuracy by week two of the quarter, attached to a SentinelOne case study. There’s no footnote on that page, no method, no sample definition. One of the seven ranking pages repeats it as though it were an independent benchmark, alongside a widely copied figure about how many Gong customers also buy Clari that traces back to a vendor blog post rather than a survey with a published method. It’s a marketing claim, and Clari is entitled to make it, but it isn’t evidence.

What I’d actually ask on the call

Ask both vendors to write the export path into the order form: which endpoint, what page size, what limit, raised to what number, for how many days after termination. Ask Gong specifically whether the 30-day deletion clock starts at contract end or at last invoice. Ask Clari which of the two APIs your conversation data lives in after the merger finishes landing, because the joint MCP server announced on July 9, 2026 is still rolling out through late summer 2026, which is now. Gong shipped its own MCP support on October 21, 2025, and its help page says it’s available on any Gong plan while noting that access depends on your assigned seat, so plan-agnostic and seat-gated at the same time. Get the seat type named.

And ask the merged entity for its subprocessor list. Google’s threat intelligence team documented a campaign it tracks as UNC6395 running August 8 to 18, 2025, which stole OAuth tokens from Salesloft’s Drift application and used them to pull Salesforce data; an August 28 update told customers to treat any and all authentication tokens stored in or connected to the Drift platform as potentially compromised. Drift is a separately acquired chatbot product that was taken offline, not Clari Forecast and not Clari Copilot. I’m not calling anything insecure. I’m saying the merged company’s subprocessor list is now the union of two previously separate lists, and asking how many third parties touch your call recordings is a countable question with a countable answer.

One more, and this is the opinion I’m least sure of: I think the merger currently makes Clari the riskier signature for a team under fifty seats, purely because packaging is in motion and the thing you’re buying may be renamed before your renewal. Ask me in October and I might have it backwards. What won’t have changed by October is the arithmetic, and the arithmetic is the part of gong vs clari that actually binds you. If you’re evaluating conversation intelligence more broadly, our Clari Copilot alternatives roundup covers the smaller end of the market. Archive first, features second.