Toutes les comparaisons · vs Recall.ai

meetbot vs Recall.ai.

Recall.ai built the category — meeting-bot-as-a-service — and the API surface they shipped is the one most of this market now copies. They have thousands of customers and public enterprise logos. We ship a Recall-style compatibility endpoint on purpose so you can test the switch in an afternoon, and then we charge you less per bot-hour. Recall's hosted API is proprietary. Our SDK + samples are MIT — clone, fork, ship, and check each third-party sample repo's LICENSE before reusing code. Below: every spec line, the math at three usage points, and a section dedicated to where they're still the better choice.

Se connecter →Aller aux tarifsvérifié le 2026-05-13

01 · résumé

La version courte.

01Recall.ai bills $0.50/hr recording, with optional built-in transcription at $0.15/hr and storage after the free retention window at $0.05/hr per extra 30 days. meetbot is a $0.30/hr entry rate for the bot-hour. Transcription on meetbot is BYOK today (Whisper/Deepgram/AssemblyAI on your key) and hosted ships Q3 2026 — so a like-for-like bundled comparison at GA, not today.
02Recall's hosted backend is proprietary. meetbot ships MIT-licensed SDKs, OpenAPI spec, and sample apps; the hosted backend is proprietary too. Hosting: Hetzner Falkenstein by default, with customer media uploaded directly to your S3 bucket.
03Recall covers 6 bot platforms today (Meet/Teams/Zoom/Webex/Slack-Huddles/GoTo) plus Desktop SDK. We expose 8 dispatch surfaces in the API (Meet/Teams/Zoom Web/Zoom SDK/Webex/Whereby/Jitsi/Discord), have zero paying customers today, and are still earlier-stage. They are the safer pick for procurement and proven scale; we are the sharper pick for price, EU default, source-available workflow, and migration control.

Utilisez Recall.ai si…

→You need Slack Huddles or GoTo coverage today.
→You need a SOC 2 Type 2 report and a HIPAA BAA in the contract on day one.
→You're indifferent to license — a proprietary hosted API is fine — and price is not a buying criterion.
→You need a mature Desktop Recording SDK today.

Utilisez meetbot si…

→Your unit economics break above $0.40/hr on the bot-hour — most LLM-on-meeting products do.
→Your customers ask where the data lives, and you'd like to answer eu-central-1.
→You may want to read the SDK + sample source under MIT and fork it without license-clause anxiety.
→You want second-metered pricing on the bot-hour, automatic volume discounts, and no storage line. Transcription is BYOK today — pipe per-speaker audio into Whisper/Deepgram/AssemblyAI on your own key (zero meetbot fee on that leg) until our hosted Whisper ships Q3.

02 · tableau des spécifications

Côte à côte. Sans esbroufe.

Chiffres vérifiés contre la source citée à la date dans le pied de page. Faites une PR de correction si quelque chose a bougé.

	meetbot	Recall.ai
recording price	$0.30 / hr entry rate with automatic volume tiers	$0.50 / hr^[1]
transcription price We do not ship transcription today. Per-speaker audio + the meeting platform's native captions are emitted; transcription is BYOK on those tracks until our hosted Whisper ships Q3.	BYOK today (free pass-through); hosted Whisper +$0.10/hr at GA (Q3 2026)	+$0.15 / hr^[2]
storage price	$0 (we don't host recordings; they upload directly to your S3)	+$0.05 / hr / 30 d^[3]
bundled $/hr (today) Apples-to-apples: 1 hr recorded + transcribed + stored 30 days. Today the BYOK path on per-speaker audio is free of meetbot fees but you pay your provider directly. At GA (hosted Whisper, Q3 2026) the bundled rate becomes $0.40/hr us vs $0.70/hr them.	$0.30 bot-hour + your own transcription bill (BYOK)	$0.70 with 30-day storage
billing granularity Recall's public pricing FAQ says Pay As You Go usage is prorated to the second; meetbot reports Stripe meter events in seconds too.	metered by the second	prorated to the second^[4]
welcome credit	$5 one-time welcome credit	5 recording hours one-time^[5]
SDK license Do not treat hosted API terms and sample repo licenses as the same thing. Check each public repo's LICENSE before reusing its code.	MIT SDK + samples; proprietary hosted backend	proprietary hosted API; public samples are repo-specific
self-hostable	planned (M5, source available); not committed	no
platforms	8 API surfaces: Meet, Teams, Zoom Web, Zoom SDK, Webex, Whereby, Jitsi, Discord	6 bot platforms: Meet, Teams, Zoom, Webex, Slack Huddles, GoTo
data residency	eu-central (Falkenstein) only	US, EU, JP^[6]
compliance certifications	none today (SOC 2 Type 1 on M5 roadmap; no BAA template; no published uptime SLA)	SOC 2, ISO 27001, HIPAA BAA^[7]
SLA	no published SLA (will publish once status.meetbot.dev has real history)	99.9% SLA on enterprise tier
transports RTMP push is not a shipping meetbot API transport today; it stays on the realtime-delivery roadmap.	signed webhooks today; optional bot-side WebSocket mirror for raw frames	webhook, websocket, RTMP
mobile SDK Recall's public product page says Mobile Recording SDK is launching soon. We do not ship mobile today either; our embedded SDK lands Q3 2026 — ambient mic capture only (Apple/Google block VoIP audio capture from other apps).	shipping Q3 2026 (iOS first)	launching soon^[8]
customers (public)	pre-launch (zero paying customers today)	thousands + public enterprise logos

[1]recording price: recall.ai/pricing
[2]transcription price: recall.ai/pricing
[3]storage price: recall.ai/pricing
[4]billing granularity: recall.ai/pricing
[5]welcome credit: recall.ai/pricing
[6]data residency: recall.ai/pricing
[7]compliance certifications: trust.recall.ai/
[8]mobile SDK: recall.ai/product/mobile-recording-sdk

03 · scénarios tarifaires

Le calcul, en trois temps.

Trois points d'usage : un hobbyiste, une startup, une entreprise à l'échelle. Formule visible par cellule — copiez-la dans un tableur, mettez vos propres chiffres.

scénario 1

Hobbyist · 10 hr / mo

10 heures d'enregistrement de réunion par mois.

meetbot

10 hr × $0.30 = $3.00 (bot-hour only)

$3.00/mois

Recall.ai

10 hr × ($0.50 + $0.15 + $0.05) = $7.00

$7.00/mois

Today: meetbot only bills the bot-hour; transcription (if you want it) is BYOK — pipe per-speaker audio into Whisper/Deepgram/AssemblyAI on your own key. Both providers' signup credits are one-time, so this returning-customer comparison does not subtract either credit.

scénario 2

Startup · 1,000 hr / mo

1,000 heures d'enregistrement de réunion par mois.

meetbot

1,000 hr × $0.30 = $300 (bot-hour) + your transcription bill (BYOK)

$300+/mois

Recall.ai

1,000 hr × ($0.50 + $0.15 + $0.05) = $700.00

$700/mois

$400/mo of saved bot-hour/storage spend at this scale — minus whatever your transcription provider charges. Whisper-large-v3 self-hosted on a Hetzner GPU box (~$200/mo for an RTX 4090) handles ~20 concurrent realtime streams; Deepgram pay-go is around $0.0043/min ≈ $0.26/hr. Real bundled cost on meetbot today: roughly $300 + $200–260 = $500–560/mo. At GA (hosted Whisper Q3 2026): $400/mo flat.

scénario 3

Scale · 50,000 hr / mo

50,000 heures d'enregistrement de réunion par mois.

meetbot

1,000 hr × $0.30 + 9,000 hr × $0.25 + 40,000 hr × $0.20 = $10,550 (bot-hour) + transcription

$10,550+/mois

Recall.ai

50,000 hr × ($0.50 + $0.15 + $0.05) = $35,000.00

$35,000/mois

At this scale you self-host Whisper-large-v3 on dedicated GPU infra (a couple of $200–400/mo Hetzner GPU boxes serve thousands of concurrent streams) and the transcription leg approaches its hardware floor. Recall negotiates volume discounts at this scale; our published self-serve rate already drops to $0.20/hr above 10k hr/mo.

04 · là où ils gagnent

Là où Recall.ai est le meilleur choix.

Nous incluons cette section parce que l'alternative — prétendre gagner partout — est malhonnête, et les pages de comparaison malhonnêtes sont la raison pour laquelle la plupart ne valent pas la peine d'être lues.

01Desktop SDK and some bot platforms. Recall ships a desktop SDK and bot support for Slack Huddles and GoTo today. We cover Webex/Whereby/Jitsi/Discord in our API surface, but the battle-tested high-volume path is still younger.
02Compliance paper. Recall has SOC 2 Type 2, ISO 27001, and signs HIPAA BAAs on the enterprise tier. We have no certifications today; SOC 2 Type 1 is on our M5 roadmap, no BAA template yet. Procurement will ask about this.
03Multi-region residency. Recall offers US, EU, and JP regions; meetbot is eu-central-1 (Falkenstein) only until first US enterprise asks for a region. APJ residency requirements are a hard miss for us today.
04Hosted transcription, today. Recall bills $0.15/hr for transcription bundled with the bot. We don't ship transcription today — BYOK only (per-speaker audio piped into your Whisper/Deepgram/AssemblyAI key, zero meetbot fee on that leg). Our hosted Whisper option ships Q3 2026.
05Mobile SDK roadmap visibility. Recall's public product page says Mobile Recording SDK is launching soon; ours ships Q3 2026 (iOS first). If mobile is the product center, both need verification before purchase.
06Volume references. They have thousands of customers and public enterprise logos — HubSpot, Calendly, ClickUp et al. — which is a real signal in procurement reviews. We have zero paying customers (we're pre-launch); that's earlier-stage risk.
07OAuth pool depth. Their hosted OAuth credential management for Zoom/Google/Teams is more battle-tested. We ship the same model in M3 but the surface is younger.

05 · là où nous gagnons

Là où meetbot gagne.

Chaque ligne renvoie à la page docs qui le prouve. Des chiffres, pas des adjectifs. Sourcé contre la surface publique de Recall.ai à la date ci-dessous.

01Lower bot-hour price ($0.30/hr vs $0.50/hr recording). At GA — once our hosted Whisper transcription lands Q3 — the bundled comparison becomes $0.40/hr (us) vs $0.70/hr (them with 30-day storage); today, BYOK transcription on per-speaker audio nets you the same gap with zero meetbot fee on the transcription leg.
preuve : /pricing
02MIT-licensed TypeScript and Python SDKs + sample apps (Go and Rust SDKs are planned). Fork them, embed them in dual-licensed products, ship them inside closed-source apps. No ELv2, no BUSL, and a clear MIT grant for our SDK and sample code.
preuve : github.com/meetbot
03EU-hosted by default. Bot containers, control plane, and database all in Hetzner Falkenstein. Recordings upload directly to your S3 bucket — no Atlantic crossing unless you point that bucket at a non-EU region.
preuve : /security
04Second-metered billing — charged by meeting time, not by container wall-clock or per-bot fee. A 17-minute call costs 8.5¢.
preuve : /pricing
05Public OpenAPI spec at github.com/meetbot/openapi — generate your own client, run it through your own SDK pipeline, audit the surface in CI.
preuve : github.com/meetbot/openapi
06Recall supports webhook/WebSocket/RTMP. meetbot's drop-in path is signed webhooks today, with an optional bot-side WebSocket mirror for raw frames; keep realtime transport differences explicit during migration.
preuve : /docs/webhooks

06 · migration

Tout le changement. Huit lignes.

Même forme, mêmes champs, hôte différent. Remplacez votre appel de dispatch de bot Recall.ai par un appel meetbot. Les payloads webhook arrivent dans la même forme JSON que votre handler parse déjà.

Recall.ai (avant)ts

// Recall.ai
const res = await fetch("https://us-west-2.recall.ai/api/v1/bot", {
  method: "POST",
  headers: {
    Authorization: `Token ${process.env.RECALL_KEY}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    meeting_url: "https://meet.google.com/abc-defg-hij",
    bot_name: "notes",
    transcription_options: { provider: "deepgram" },
    real_time_transcription: { destination_url: WEBHOOK_URL },
  }),
});

meetbot (après)ts

// meetbot — same shape, different host.
// Note: we don't ship transcription today — per-speaker audio
// lands in your bucket; pipe it into your provider of choice.
// Hosted Whisper option lands Q3 2026.
const res = await fetch("https://api.meetbot.dev/api/v1/bot", {
  method: "POST",
  headers: {
    Authorization: `Token ${process.env.MEETBOT_KEY}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    meeting_url: "https://meet.google.com/abc-defg-hij",
    bot_name: "notes",
    webhook_url: WEBHOOK_URL,
    real_time_transcription: { destination_url: WEBHOOK_URL },
  }),
});

07 · faq

Les questions qu'on nous pose vraiment.

Q.How does the price actually compare?

Today: $0.30/hr (us, bot-hour only) vs $0.50/hr (Recall, recording only). Bundled with transcription + 30-day storage Recall is $0.70/hr; we don't bundle transcription today (BYOK on per-speaker audio is free of meetbot fees, you pay your provider directly). Once our hosted Whisper ships Q3 2026 the apples-to-apples bundled comparison becomes $0.40/hr (us) vs $0.70/hr (them with 30-day storage). The pricing scenarios above show today's math; the calculator at /pricing lets you plug in your own assumptions.

Q.How long does the migration actually take?

Most teams can smoke-test in an afternoon. The HTTP shape is intentionally familiar — we ship POST /api/v1/bot as a Recall-style create-bot shim for common snake_case payloads, then translate internally to POST /api/v1/jobs. Webhook payloads are close, but you should still run your parser against our fixture before switching production traffic.

Q.What if I need Webex or Slack Huddles?

Webex is in our API surface now, but if you need Slack Huddles or GoTo, stay on Recall today. Discord is our wedge in the opposite direction: it uses the native Discord voice gateway instead of browser automation.

Q.Do you sign a BAA?

No. We have no BAA template today, no SOC 2, no HIPAA posture. SOC 2 Type 1 is on the M5 roadmap; the BAA template would follow that audit. If you have a HIPAA-bound deployment timeline, this is a real reason to stay on Recall today and re-evaluate us once those land.

Q.Where is the data physically stored?

Bot containers run on Hetzner Falkenstein (DE). Postgres + recording manifests live on the same host. The recording media itself goes to whatever S3-compatible bucket you nominate — if you point that at AWS Frankfurt, the entire pipeline stays in EU borders. If you point it at AWS us-east-1, only that final hop leaves the EU. We can send a DPA on request (template-stage, not a published certified document).

Q.What about the Recall.ai SLA?

Recall offers a 99.9% SLA on their enterprise tier. We don't publish an uptime number or a contractual SLA — we will once status.meetbot.dev is up with real history. If contractual uptime is a buying criterion today, that's a real reason to stay on Recall.

Q.How open is the source, exactly?

The TypeScript and Python SDKs, CLI, sample apps, and OpenAPI spec are MIT — clone, fork, ship. The bot binary itself stays closed in our monorepo (apps/bot) — the same Vercel/Next.js shape. We may publish a source-available bot in M5 under a permissive license; not committed yet. By contrast, Recall's hosted API is proprietary; check each public sample repo's LICENSE before reusing its code.