Sesame logo
Personal Assistants · Voice Assistants

Sesame

Sesame is a set of voice-first personal agents you talk to like a person. Four characters hold real-time spoken conversations, search the web, take notes and remember what you discussed, on iPhone, Android and in a browser.

Beta GDPR compliant Free plan Free No public API 18+ Verified by Guidaio
Overview

What is Sesame?

Sesame is a collection of voice-first personal agents built by Sesame AI, Inc., a San Francisco company whose stated ambition is to make speaking to a computer feel closer to speaking to a person. Four characters are available in the current preview: Maya, warm and creative; Miles, laid-back and sharp; Simone, curious and intellectual; and Charlie, witty and warm. Each has its own voice, personality and point of view, and choosing between them is presented as part of personalising the experience.

The product runs on three surfaces. The Mobile Preview is on iOS and Android; the Research Preview runs in a browser. On mobile, or signed in on the web, agents converse by voice and text, search the web, run Deep Dives for longer answers, create notes and reminders, and send a summary once a call ends. The browser Research Preview is deliberately narrower: voice calls and web search only. Calls run up to 30 minutes when you are signed in and 5 minutes when you are not, and can be restarted immediately.

Memory sits at the level of the individual agent. What you discuss with Miles stays with Miles; Maya keeps her own thread; they do not compare notes. That memory follows you between web and mobile so long as the account is the same. Incognito Mode inverts the arrangement, letting agents draw on earlier context while writing nothing new to memory or to Sesame's servers.

Two integrations are documented. Gmail lets an agent search and read mail, draft and send messages with your approval, and work with labels and contacts. Google Calendar covers viewing, creating, updating and deleting events, finding mutual availability across shared calendars and setting up recurring meetings. Both are installed by asking an agent for them and authenticating with Google.

Behind the product sits Sesame's own research. The company published the Conversational Speech Model, a single-stage end-to-end transformer approach trained on roughly one million hours of mostly English audio, and open-sourced key components. Its goal, in its own words, is voice presence: emotional intelligence, conversational timing, contextual awareness and a consistent personality. English remains the only supported language, and Sesame states plainly that its agents sometimes invent things, including facts about Sesame itself.

What it does

  • Hold a real-time spoken conversation with an agent whose voice and personality you choose
  • Search the web mid-conversation, running several queries in parallel and folding results in as they arrive
  • Run a Deep Dive when a question deserves a longer, more thorough answer
  • Capture notes and set reminders by voice, then receive a written summary after each call
  • Switch to text with the same agent, keeping one shared memory across speaking and typing
  • Search, read, draft and send Gmail messages with your approval, and manage labels and contacts
  • View, create, update and delete Google Calendar events, find mutual availability and set recurring meetings
Audience

When to use Sesame / When not to

A quick filter to help you decide if Sesame is the right fit.

When to use Sesame

  • Curious people who think best out loud, on a commute or an evening walk, and want a spoken thinking partner rather than a prompt box
  • Founders, consultants and freelancers who need to talk a decision or a half-formed project through before committing to it
  • Students and early-career job seekers using Deep Dives and web search to explore a subject and capture what matters as notes
  • Executive assistants and office managers who want an agent to search, draft and send Gmail messages and juggle Google Calendar by voice
  • Speech and conversational-AI researchers, who can read Sesame's published work on the Conversational Speech Model and use the open-sourced components

When not to use Sesame

  • Anyone seeking legal, financial, medical, tax or other professional advice: the Acceptable Use Policy forbids it outright
  • People in distress or in an emergency, since the service is explicitly not a crisis, therapy or emergency-care tool
  • Non-English speakers, because English is the only officially supported language
  • Developers and teams wanting to embed the technology, as there is no public API and no developer documentation
  • Businesses with procurement requirements, given the absence of a published DPA, subprocessor list or security certification
Get started

How to use Sesame

A typical end-to-end flow, from setup to results.

  1. Open the Research Preview in a browser at app.sesame.com, where calls of up to five minutes need no account
  2. Or install the Mobile Preview from the App Store or Google Play
  3. Create an account or sign in; a short waitlist may follow sign-up during the rollout
  4. Pick an agent from the selector at the top of the screen, previewing Maya, Miles, Simone and Charlie to find the right fit
  5. Start a call, or send a text message if speaking aloud is awkward
  6. Open with something broad rather than a polished prompt: a question you keep circling, a project you want to start, a decision you are weighing
  7. Ask the agent to save anything worth keeping as a note, and to set reminders as they come up
  8. Connect Gmail or Google Calendar by asking the agent to install the app, then authenticate with Google; connected apps live under Settings, Integrations
  9. Turn on Incognito Mode for a conversation that leaves no trace, and use account settings to opt out of model training
  10. Report bugs by shaking your phone, or through the post-call form; delete everything from Settings, your name, Delete my account
Quick read

Pros & Cons

Pros

  • Free during the preview, with the full experience rather than a cut-down tier
  • Available on iOS, Android and the web, with memory carried across them
  • Conversation quality is the explicit engineering focus, down to pivoting mid-sentence as search results arrive
  • Incognito Mode is genuinely ephemeral, and account deletion is self-service from Settings
  • Training opt-out available; unregistered web users are never trained on without consent and their data is deleted within thirty days
  • Data pulled from Gmail and Calendar is never used to train the models, and is encrypted in transit and at rest
  • Unusually open for the category: published research, open-sourced model components, and a named leadership team and investor list

Cons

  • Still a preview, with a waitlist and a deliberately narrow set of capabilities
  • English is the only officially supported language, and mobile availability is limited to 39 unnamed countries
  • Calls are capped at 30 minutes signed in and 5 minutes otherwise
  • No public API, no developer documentation and no published pricing, so the post-preview business model is unknown
  • For signed-in users, training on conversations is the default and the opt-out must be switched on
  • The licence granted over your Content is perpetual and irrevocable, and survives deleting it or closing the account
  • No DPA, subprocessor list, standard contractual clauses or security certification, with all hosting in the United States and mandatory individual arbitration
Pricing

Pricing & Plans

Sesame publishes no pricing page and no paid tier. During the current rollout the company states that the full experience is free, on mobile as on the web, subject to a possible short waitlist after sign-up. No amount, currency or billing unit is therefore available. The Terms of Use do refer to sums a user may have paid and to fees that are not refundable, which leaves room for a paid offering later, but nothing of the sort is announced today.

Plan 1
  • Mobile Preview (iOS and Android) — free during the rollout
  • account required
  • calls up to 30 minutes
Plan 2
  • Research Preview (web
  • app.sesame.com) — free
  • usable without an account for calls up to 5 minutes
  • up to 30 minutes signed in
Prices and plans listed above may evolve. Always check the official pricing page before subscribing.
Trust & Privacy

Data, GDPR & hosting

A consolidated view of how Sesame handles your data.

GDPR overview

Section 16 of the privacy policy is written for residents of the EEA, Switzerland and the United Kingdom. It lists access, rectification, erasure, objection, restriction, portability and withdrawal of consent, promises a reply within one month, and points to the national supervisory authority for complaints. A lawful basis is stated purpose by purpose, and consent is collected before any non-essential cookie. A GDPR representative and Data Protection Officer is reachable at legal@sesame.com, though no organisation or individual is named. Sesame warns that once data has fed a model it can no longer identify you inside it, which may limit what it can act on. No Article 27 representative, no standard contractual clauses and no DPA are published, and the service is declared operated in and intended for the United States.

Who owns the data?

You keep ownership of your Content, meaning both the Input you provide and the Output the agents return; Sesame's own models, algorithms and technology are excluded from it. In exchange you grant Sesame a perpetual, irrevocable, non-exclusive, worldwide, royalty-free licence to host, use, modify and create derivative works of that Content in order to run and improve the service, and, for signed-in users, to train its models unless you opt out in your account settings. That licence outlives deletion of the Content and closure of the account. Sesame may share Content with staff and advisers, service providers, authorities, group companies and any future acquirer, but commits not to commercialise your voice on a standalone basis without permission.

Reuse rights

The licence you receive is personal, non-exclusive, non-transferable and revocable, and covers your own use only. You may generally post what you produce to social media or livestream your sessions, provided you follow the Terms and the brand guidelines and label the material as AI-generated. Beyond that the restrictions are firm: no commercial use of the service, no comparative analysis, no using Output to build a competing product, and no using it to create, train, test or improve any other AI model. Output about an identifiable person must never be used in a way that affects them legally or otherwise seriously. Sesame also warns that Output is not unique, so another user may receive something very similar.

Data retention & training

Retention summary
Sesame keeps personal information only while it has a legitimate business reason to, such as completing the purpose of collection, maintaining the relationship, resolving disputes, or meeting legal, tax and accounting obligations. Unregistered web users are treated more protectively: their data is deleted automatically within thirty days and is never used for training without consent. For signed-in users, Sesame reserves the right to retain content for as long as the law allows, and closing or suspending an account does not oblige it to delete anything. You can delete your account and its data yourself from Settings, or write to support or legal. Data that has been de-identified or aggregated is no longer treated as personal and may be reused, including for training. Incognito Mode saves nothing at all.
Trains on customer data
Configurable
Training opt-out available
Yes
GDPR contact

Hosting summary

Sesame operates its services in the United States and describes them as intended for users located there. If you are outside the country, the privacy policy is explicit that your information will be transferred to, processed, stored and used in the United States, that data protection law there may differ from your own, and that your data may be subject to access requests from United States governments, courts and law enforcement. It adds that data may sit on servers outside your jurisdiction, in places with less stringent privacy practices. No other hosting country is mentioned and no region, data centre or hosting provider is named. On security, Sesame commits to reasonable technical, administrative and physical safeguards, and states that data retrieved from Gmail and Calendar is secure in transit and encrypted at rest. No SOC 2, ISO 27001 or equivalent certification is claimed, and no subprocessor list is published.

Hosting countries
🇺🇸 United States
Availability

Where Sesame works

Country-level availability.

Not available in

No excluded country or region is named anywhere on the site. The iOS preview is described as available in 39 countries, but those 39 are not listed; the Terms state that the services are operated in and intended for the United States, and the export Control clause bars use from countries under United States embargo and by parties on United States restricted Party lists, again without naming any
Watch-outs

Things to keep in mind

Risks and trade-offs to weigh before adopting Sesame.

  • Emotional attachment: agents have names, voices, personalities and memory, and are designed for the intimate in-between moments of the day. Romantic roleplay is barred, but a parasocial bond is not
  • Sesame states that its agents sometimes make things up, including about its own product, team and roadmap, and that only the website is a source of truth
  • Cognitive outsourcing: a partner that thinks out loud with you can quietly become a partner that thinks instead of you
  • Privacy of speech: voice recordings and transcripts are collected, and for signed-in users training on them is the default rather than an opt-in
  • Connecting Gmail and Calendar exposes genuinely sensitive material, which Sesame itself warns about before you authorise it; staff may in rare cases review conversations, and inputs may be sent to third-party moderation tools
  • The licence over your content is perpetual and irrevocable and outlives deletion of the content and closure of the account
  • Limited recourse: mandatory individual arbitration, a class-action and jury waiver, liability capped at the greater of 100 USD or twelve months of payments, and all data processed in the United States
Setup

Setup & Integrations

Technical difficulty

Essentially none. You either open a web page or install an app from the App Store or Google Play. The browser version works without an account for short calls; an account unlocks memory, longer calls and integrations, though a brief waitlist is possible. Connecting Gmail or Google Calendar means asking an agent to install the app and authenticating with Google, with no configuration of your own. All you need is a microphone and an internet or mobile data connection. There is no API key, SDK or developer documentation to deal with.

Deployment

Web appMobile appIOS appAndroid app

Apps stores

Integrations

Gmail Google Calendar

Supported languages

English
Company

Behind Sesame

Company name
Sesame AI, Inc.
Founded
18/11/2022
Country of origin
🇺🇸 United States
Headquarters
150 Spear St, Suite 1800, San Francisco, CA 94105
UBO
INFORMATION_NOT_FOUND
UBO country
INFORMATION_NOT_FOUND
Domain registrar country
🇺🇸 United States
Legal contact
Support contact

Fundraising

Sesame states on its team page that it is backed by a16z, Sequoia, Spark, Matrix and a group of founders and investors
A Form D filed with the SEC on 12 September 2023 records 10,124,975 USD placed with 22 investors, first sale dated 31 August 2023
Series A of 47.5 million USD led by a16z, announced on 27 February 2025 (press reports, not stated on the site)
Series B of 250 million USD led by Sequoia Capital and Spark Capital, announced in October 2025, bringing reported cumulative funding to roughly 307.6 million USD (press reports, not stated on the site)

Social

Official links

Resources

All the official URLs gathered for verification and reference.

FAQ

Frequently asked questions

What exactly is a Sesame agent?
A personal agent built for conversation and collaboration. Four are offered in the preview — Maya, Miles, Simone and Charlie — each with its own voice, personality and point of view. They are made for following your curiosity, exploring an idea or talking through what is on your mind.
Where can I use it, and what does each version do?
On iOS and Android through the Mobile Preview, and in a browser through the Research Preview at app.sesame.com. On mobile, agents handle voice and text, web search, Deep Dives, notes, reminders and post-call summaries. The web version is limited to voice calls and web search.
How much does it cost?
Nothing at present. Sesame states that during this phase of the rollout the full experience is free. There is no pricing page and no paid plan, though you may hit a short waitlist after signing up.
How long can a conversation last?
Up to 30 minutes when you are signed in or using the mobile app, and 5 minutes if you are not signed in on the web. You can restart a call immediately once it ends.
How does memory work?
Each agent keeps its own memory, so what you tell Miles stays with Miles and Maya keeps her own thread. Memory follows you between web and mobile while you use the same account. In Incognito Mode agents can still draw on past memories but create no new ones.
Are my conversations used to train the models?
For signed-in users, yes by default, unless you opt out of training in your account settings. Unregistered web users are not trained on without consent, and their data is deleted automatically within thirty days. Data retrieved from Gmail or Calendar is never used for training.
Can I delete my data?
Yes. Go to Settings, tap your name and choose Delete my account at the bottom of the screen. Transcript requests and data deletion enquiries can also be sent to legal@sesame.com.
Which languages and integrations are supported?
English is the only officially supported language. Two integrations are documented, Gmail and Google Calendar; you install them by asking an agent, then authenticating with Google, and manage them under Settings, Integrations.
Is there an API for developers?
No. Sesame publishes no public API and no developer documentation. It has, however, open-sourced key components of its Conversational Speech Model research on GitHub.
Who is behind Sesame, and is there an age limit?
Sesame AI, Inc., headquartered in San Francisco with offices in Bellevue and New York, founded by Brendan Iribe, Ankit Kumar and Ryan Brown and backed by a16z, Sequoia, Spark and Matrix. You must be 18, or older where local law requires it.
Conclusion

Should you pick Sesame?

Sesame is one of the more coherent bets on voice as a primary interface rather than a novelty layer. The engineering effort visibly goes where it matters for that bet: latency, conversational timing, the ability to keep talking while searches land, and a memory that makes each character feel like the same presence a week later. The published research behind it, and the decision to open-source parts of the Conversational Speech Model, give the claims more substance than a marketing page would.

The scope is narrow on purpose. Agents converse, search, dig deeper, take notes, set reminders, and reach into Gmail and Google Calendar. They do not code, do not integrate with anything else, and are explicitly barred from professional advice. English is the only supported language, mobile reach stops at 39 countries, and calls end at 30 minutes. Anyone hoping to build on the technology will find no API.

The trade-offs are worth reading closely. Everything is free today, with no published pricing and no signal about what happens next. For signed-in users, training on conversations is the default rather than a choice you make, and the opt-out sits in settings you have to find. The licence granted over your content is perpetual and survives deleting it. Data lives in the United States, no DPA or subprocessor list is published, and disputes go to individual arbitration with liability capped at 100 USD. Sesame is also unusually candid that its agents invent things, including about Sesame.

Taken together: a genuinely pleasant, well-made conversational companion for thinking out loud, best approached as a free preview from a well-funded company still deciding what it will charge for — and worth entering with your privacy settings reviewed first.