Sayso listens to both sides of your call and puts the answer on your screen in under a second. Not two. Not four. Under one.
Free to start. No card. Your email is the whole sign-up.
Mac (macOS 13+) · Windows soon · Web dashboard
Sayso reads system audio directly, so it hears your customer as clearly as it hears you. No bot in the participant list, no “recording this call” banner, no awkward explanation at the top of the meeting.
system audio · diarisedDana · Acme
System audioYou
MicrophoneDanaAnd how long does onboarding usually take for a team like ours?
YouGood question. For a team your size
Your playbook, price list and the CRM record for this account load once when the call starts and stay cached for the whole session. That is why answers are specific, and it is half the reason they are fast.
cached 8ms · reused every turnOnboarding, section 4
Seat bands, row 14
Acme Corp · 40 seats
About two weeks for a team of forty, with a pilot group first.
Source: Onboarding guideA discreet panel that never takes keyboard focus, never moves your cursor, and never appears in a screen share. Your customer sees the call. You see the call and the answer.
setContentProtection · never takes focus · never moves your cursorThe panel is excluded from capture by the operating system, so a screen share carries the call and your deck, and nothing else. On every plan, not a stealth tier.
Plans, 2026
Per seat, per month, billed annually
Your screen: the call, your deck, and the Sayso panel over them.
Sayso loads your context before the first hello, answers during, and closes the loop after: notes, follow-up, CRM.
Pick the account. Playbook, price list and the CRM record load once and stay cached for the whole call.
“Yes, 1-hour first response on P1, in the Enterprise tier.”
Every answer names its source, a playbook section, a price-book row, a CRM field, so you can say it with a straight face.
Action items, decisions and owners, pulled from what was actually said rather than what you remember.
Thanks for the time today. As promised: about two weeks for a team of forty, with a pilot group first. The band table is attached.
A send-ready email that quotes the numbers you actually gave, waiting when you hang up.
Fields updated and the next step booked, without the twenty minutes of admin.
Every objection your team hears becomes an answer the whole team gets, next call.
“Unlike the incumbent, pricing is per active seat, and you keep your integrations.”
Playbook · competitors“How is this different from what we already have?”
The competitor section of your playbook, surfaced the moment they name the incumbent.
“Overages bill at the same rate, prorated to the day. Or pre-buy a band at 15% off.”
Price book · overage“What if we go over the seat count mid-contract?”
Overage terms and band pricing from your price book, prorated the way your finance team wrote it.
“SSO and SAML on Team. EU or US residency, pinned per workspace.”
Security docs · residency“Does it work with our SSO and what about data residency?”
The security answers from your docs, verbatim, so you never round a compliance detail.
“Twelve seats were added on the 9th, prorated to the day. That is the whole change.”
CRM · Acme invoices“Why did our invoice change this month?”
The account record and the last three touchpoints, on screen before you have to say “let me check”.
“Three findings. Churn concentrates in month two, onboarding is the lever, the fix is cheap.”
Audit deck · findings“What did you find in the audit, in one sentence?”
Your own findings and the client’s numbers, loaded from the deck you sent them yesterday.
“Sent: the notes, three action items and the band table. Two minutes ago.”
This call · notes“Can you send that over after?”
Notes, action items and a draft email from what was actually said. Ready when you hang up.
The tool is not there to think for you. It is there so the two seconds you spend remembering a number do not cost you the room.Why we build for work calls, not exams
Salesforce
HubSpotSlack
Notion
Google CalendarSalesforceHubSpotSlackNotionGoogle CalendarCall platforms in, CRM and notes out. Everything else through webhooks and the REST API.
Reads system audio at the OS level, stays out of screen shares, never takes focus.
A synthetic call runs through the real pipeline every hour. Live p50, uptime and per-platform results on /status.
Competitor figures are the best published or independently reported values we could find. If any of them is wrong, tell us and we will correct it here.
| Sayso | Parakeet AI | Cluely | LockedIn AI | |
|---|---|---|---|---|
| Median time to first token | 684ms, published | 2,000 to 4,500ms | 5 to 10s reported | not published |
| Live status page | Yes | |||
| Publishes a latency number | Yes | |||
| Speech provider failover | 2, hot | 1 | 1 | 1 |
| Discreet panel on every plan | Yes | Yes | $149.99/mo tier | Yes |
| CRM write-back | Yes | none | none | none |
| Team workspace, SSO, audit log | planned | none | none | none |
| Entry price | free, then $30 | $25 one-off | $44.99/mo | $29.99/mo |
No audio kept, no training on your data, and a second speech provider standing by. The rest is in the docs.
Audio is streamed, transcribed and discarded. Transcripts are opt-in per workspace, and your playbook and CRM context are cached for the session only.
Choose EU or US processing per workspace. Speech and model providers are pinned to that region, and the sub-processor list is published and versioned.
Two speech providers, hot. If a session still fails technically, it is refunded without you asking and the incident shows on the status page.
SSO, SAML, role permissions and an audit log are planned for teams. Full detail in the security docs.
Cluely charges $149.99 a month to stop your overlay showing in a screen share. We think that should be the product, not an upgrade.
Try it on real calls: 5 calls, 10 minutes each, no card.
For a stretch of calls back to back. Or $399 a year.
Pay for the hours you use: 3 credits are 3 hours of calls, used in 30-minute steps.
Straight answers. Fitting, really.
Anything else, the docs go deeper on the audio pipeline, retention and the API.
No. Sayso reads system audio on your machine, so nothing appears in the participant list and no recording banner is triggered. The other side sees a normal call.
From the moment the other person stops talking to the first word painted on your screen, on a scheduled synthetic call through the real pipeline. The budget: 300ms to detect they finished talking, 8ms to assemble context that was already cached, 350ms for the model's first token, 26ms to paint. Live p50 and p95 are published on /status.
No. The panel is excluded from capture at the OS level, never takes keyboard focus and never moves your cursor. It is part of every plan, not a separate tier.
Speech-to-text fails over to a second provider automatically. If a session still fails, it is refunded without you asking.
SSO and SAML, role-based permissions, retention controls and an audit log are planned for teams. Today, audio is discarded after transcription and nothing from your calls is used for training.
No. Nothing from your calls is used for training, by us or by any provider we route to. Audio is discarded after transcription and context is cached for the session only.
No bot joins. They see a normal call.
Cancel in one click. Failed sessions refund themselves.
Live latency and uptime published on /status.
