Recall.ai
Recall.ai is a developer API and SDK suite for capturing meetings. It returns recordings, speaker-labelled transcripts and metadata from Zoom, Google Meet, Teams, Webex and more, either through a meeting bot or through bot-free desktop recording.
What is Recall.ai?
Recall.ai is an infrastructure layer rather than a finished product. Published by Hyperdoc Inc. of San Francisco, it sells software companies a single interface for capturing what happens inside meetings, so that those companies can build their own features on top instead of maintaining recording pipelines of their own.
Four products sit behind that promise. The Meeting Bot API sends a visible bot into the call; the vendor recommends it wherever explicit recording consent matters, or where an AI agent needs to be present in the room. The Desktop Recording SDK takes the opposite approach, capturing from the user's own machine with no bot in the meeting, at the cost of requiring that user to install a lightweight application and to attend the call. The Calendar API connects Google Calendar and Microsoft Calendar, dispatching bots from the diary and enriching recordings with participant emails and meeting titles. A Mobile Recording SDK is announced but still marked Coming Soon.
Coverage spans Zoom, Google Meet, Microsoft Teams, Webex, GoTo Meeting and Slack Huddles, with more than a hundred distinct pieces of meeting data exposed through one unified API. The technical argument the company makes for itself is that it captures each participant's audio separately rather than reconstructing who spoke afterwards, which it says yields exact diarisation, and that its video output looks like a genuine recording rather than a screen capture cluttered with interface elements and notifications.
Transcription can come from Recall.ai's own engine or from third-party providers such as Deepgram, AssemblyAI, Speechmatics, AWS Transcribe, ElevenLabs and Rev, using the customer's own keys. Data is held in one of three isolated regions: US West, the European Union or Asia.
The customer list is public and substantial, naming ClickUp, Apollo.io, Incident.io, BrightHire, Filevine and Mindtickle among others, with more than 3,000 customers claimed. The stated security posture covers SOC 2 Type II, ISO 27001, HIPAA with a business associate agreement on the upper tiers, GDPR and CCPA. A 38 million dollar Series B was announced in September 2025.
What it is not is a tool for an end user: without a developer, there is nothing here to switch on.
What it does
- Record video and audio from a call through a bot that joins the meeting
- Capture meetings from a desktop application with no bot present in the call
- Retrieve speaker-labelled transcripts, asynchronously or in real time
- Pull separate audio and video streams for each individual participant
- Enrich recordings with calendar metadata, meeting titles and participant emails
- Send and read meeting chat messages, and stream audio or images into the call
- Detect, enter and record breakout rooms
When to use Recall.ai / When not to
A quick filter to help you decide if Recall.ai is the right fit.
When to use Recall.ai
- Product and engineering teams adding meeting recording to an existing SaaS product without building the capture infrastructure themselves
- Builders of AI notetakers, sales conversation intelligence tools and CRM automation that need clean, speaker-labelled transcripts
- Early-stage startups validating a meeting-data product, drawn by the five free hours and the discounted startup rate
- Enterprises with data residency obligations, who need recordings held in the US, the EU or the APAC region
- Teams building interactive AI agents that must join a call, speak and display media rather than merely listen to it
When not to use Recall.ai
- End users looking for a ready-made notetaker: Recall.ai sells the plumbing, not a finished meeting assistant
- Teams with no developer on hand, since every product is consumed through an API or an SDK and no no-code interface exists
- Anyone needing a phone app today, as no iOS or Android application is published and the Mobile Recording SDK is still marked Coming Soon
- Buyers who need a permanent free tier: past the first five hours, every minute recorded is billed
- Organisations that must fix a budget upfront, because only the entry plan carries a published price
How to use Recall.ai
A typical end-to-end flow, from setup to results.
- Create an account through the sign-up page, choosing the region that will hold your data
- Collect the API key issued for that specific region from the dashboard
- Point every request at the base URL of your region, whether us-west-2, eu-central-1 or ap-northeast-1
- Work through the quickstart in the documentation, which claims a production integration in around twenty-four hours
- For the bot route, send a POST request carrying the meeting URL to create and dispatch a bot
- For the bot-free route, embed the Desktop Recording SDK in your desktop app, built with Electron, Tauri or native macOS and Windows code
- Handle the macOS permissions for microphone, screen recording and accessibility; Windows requires none
- Subscribe to webhooks for bot status changes and for real-time audio and video events
- Connect the Calendar API if you want bots dispatched automatically from users' diaries
- Call the Retrieve Recording endpoint to collect audio, video, transcripts and metadata once the meeting ends
Pros & Cons
Pros
- One integration covers six conferencing platforms, Slack Huddles included through the desktop route
- Two capture models, visible bot or no bot at all, sit behind the same API and cost the same
- Data residency across three isolated regions, which the company claims no competitor matches
- Substantial compliance posture: SOC 2 Type II, ISO 27001, HIPAA, GDPR and CCPA
- Billing is prorated to the second and indifferent to the number of participants in the call
- Generous on-ramp: five free hours, then a startup rate of 0.25 dollars per hour for the first 10,000
- Named public customer references with published case studies, backed by a 38 million dollar Series B
Cons
- Only the entry tier carries a published price; Launch and Enterprise are both quoted individually
- No permanent free plan, since the five free hours are an on-ramp rather than a tier
- The site says nothing at all about whether customer data is used to train models
- True cost is hard to forecast, as transcription and retention are billed on top of the hourly rate
- The Mobile Recording SDK is advertised across the site but is not actually available
- The competitive benchmark presented as objective is produced by Recall.ai about its own rivals
- No about page, no contact page, and the trust portal does not render without JavaScript
Pricing & Plans
There is no permanent free plan. Access begins with five hours of recording offered free of charge on the Pay As You Go tier, after which recording is billed at USD 0.50 per hour, prorated to the second and independent of the number of participants. Built-in transcription adds USD 0.15 per hour, and storage beyond the seven free days costs USD 0.05 per hour of recording retained for thirty days. The Launch and Enterprise tiers are quoted individually, with volume discounts negotiated through the sales team.
- for early-stage startups building a proof of concept - USD 0.50 per hour of recording
- with the first five hours free
- for companies launching their product - custom pricing
- adding white-glove support over Slack
- SSO
- uptime guarantees
- custom contract redlining and a HIPAA business associate agreement
- for large companies wanting white-glove security
- support and infrastructure - custom pricing
- with the same additions as Launch
Data, GDPR & hosting
A consolidated view of how Recall.ai handles your data.
GDPR overview
Implementation here is concrete rather than declarative. A full EU and UK Data Processing Addendum is published and freely readable, incorporating the European Commission's Standard Contractual Clauses 2021/914 of 4 June 2021 together with the UK Addendum, under Modules Two and Three. An Article 27 representative is formally designated: Instant EU GDPR Representative Ltd, Adam Brogden, reachable at contact@gdprlocal.com in Lucan, County Dublin. An internal data protection officer is named, David Gu, with a direct address and telephone number, and the eight data subject rights are listed explicitly. Transfers to the United States are acknowledged as lacking an adequacy decision and are covered by Article 46 safeguards. The site never writes the phrase GDPR compliant, but the apparatus standing behind that claim is unusually complete.
Who owns the data?
Under the published Data Processing Addendum, the customer acts as controller or processor and Hyperdoc Inc. is the processor: it may handle personal data only for the purposes set out in the agreement and on the customer's documented instructions. The customer remains solely responsible for the accuracy, legality and provenance of whatever is sent in. Recall.ai commits in its privacy policy not to sell personal information, and takes service-provider status under the CCPA, which bars any resale. When the service ends, personal data is returned or deleted at the customer's choice, and a certificate of deletion is issued on request.
Reuse rights
Processing is confined to running the service: capturing meetings, transcribing them and delivering media and metadata back to the customer. The Addendum separates two further streams the vendor handles in its own right, namely account data covering billing and authorised users, and usage data covering logs, performance and abuse prevention. Sharing is limited to sub-processors that help deliver the service, and the list of those sub-processors is published on the trust portal. One silence deserves flagging: nowhere across the pages collected does Recall.ai state whether customer data is used to train models, in either direction.
Data retention & training
Hosting summary
Recall.ai runs three data-isolated regions, each with its own base URL, its own sign-up route and its own API keys: US West on us-west-2, the European Union on eu-central-1 and Asia on ap-northeast-1. The pricing page frames this as a choice between US, EU and JP data residency, and workspaces can be migrated between regions after the fact. The comparison page claims Recall.ai is the only provider in its category to offer all three. Recording storage is itself optional. The platform can hold media on your behalf, or pass it straight through to your own infrastructure where compliance requires it. Separately from recording data, the privacy policy states that personal data about website visitors and customers, meaning names, job titles, employers and work contact details, is stored in the United States on databases hosted by third parties. Transfers into the United States are acknowledged as falling outside any adequacy decision and are covered instead by Standard Contractual Clauses.
Things to keep in mind
Risks and trade-offs to weigh before adopting Recall.ai.
- Recording a conversation is a legal act, and the duty to obtain participants' consent falls on the product you build rather than on Recall.ai, with rules that differ by jurisdiction
- Media attached to a recording is kept indefinitely by default; zero retention is available but must be deliberately configured
- Nothing on the site states whether customer recordings feed model training, which on conversational data is the question that matters most
- The HIPAA business associate agreement is confined to the Launch and Enterprise tiers, so a healthcare prototype running on Pay As You Go is not covered
- Bot-free desktop capture is quieter by design, which makes it easier to record people who have not registered that recording is under way
- Zoom price increases are passed through to you under the terms, capped at five percent a year
- The registered address is a mail box on Market Street rather than an identifiable head office, which limits what can be verified about the entity
Setup & Integrations
Technical difficulty
Developer-only. There is no interface to click through: you create an account, pick a region, take an API key and call the endpoints. The company advertises a production integration in twenty-four hours and a couple of lines of code, which is fair for the bot route. The Desktop Recording SDK is a heavier lift, since it must be embedded in your own desktop application and macOS permissions for microphone, screen recording and accessibility have to be handled. Documentation is thorough. Without a developer, the product cannot be used at all.
Deployment
Integrations
Supported languages
Behind Recall.ai
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement Recall.ai.
Frequently asked questions
Can I try Recall.ai for free?
Does the number of people in a meeting change the price?
Which meeting platforms are supported?
What is the difference between the Meeting Bot API and the Desktop Recording SDK?
Where is my data stored?
Is there a mobile app?
Does the Desktop Recording SDK listen all the time?
What certifications does Recall.ai hold?
Is there a minimum age to use the service?
How do I get support?
Should you pick Recall.ai?
Recall.ai occupies a clear and defensible position: it is the plumbing beneath a whole category of meeting-AI products, not something a meeting participant would ever open. That focus shows in the quality of the offer. Platform coverage is the broadest in its peer group, the choice between a visible bot and silent desktop capture is genuinely useful rather than cosmetic, and the three isolated data regions answer a procurement question that stops many deals dead.
The compliance work is real. A published Data Processing Addendum carrying the current Standard Contractual Clauses, a named Article 27 representative, an internal data protection officer with a working address, SOC 2 Type II and ISO 27001: this is more than most vendors of this size bother to assemble, and it is the strongest single reason to take the company seriously.
Two reservations temper that. The first is commercial. Only the entry tier shows a price, and the headline 0.50 dollars per hour is not the whole bill once transcription and retention are added. Anyone modelling costs at scale will have to talk to sales, which is precisely the moment when leverage is lowest. The second is informational. The site is entirely silent on whether customer recordings are used to train models. That silence is not evidence of misuse, but on a product whose raw material is private conversation, it is a question a buyer should put directly rather than infer.
Read the published competitive benchmark with the same caution: it is Recall.ai measuring its own rivals, and it unsurprisingly wins.
For a technical team that needs meeting capture and would rather not spend a year discovering the edge cases first-hand, the case is strong. For anyone without a developer, there is nothing here to use.
- Choosing a selection results in a full page refresh.
- Opens in a new window.