Conva
Real-time call co-pilot for sales reps — dual-channel live transcription and AI-suggested replies with document citations, processed locally on your desktop.
Conva is a real-time call intelligence tool, not a receptionist — it doesn't answer the phone in place of a person. Instead it sits alongside a sales rep who's already on a call: dual-channel capture keeps the rep's voice and the other party's separate, live speaker-labeled transcription lands in under a second, and an AI assist layer suggests replies grounded in the rep's own uploaded documents, citing the specific source passage rather than answering from a black box. Calls are recorded and searchable afterward.
Its privacy story has two layers worth telling apart. Capture and speech-to-text run fully on-device via an offline model — per Conva, audio never leaves the computer, and nothing is stored on its servers unless explicitly saved. The AI-suggested-replies feature is a separate, opt-in cloud path: Conva describes it as bring-your-own-key ("your keys, your provider") for the cloud speech/LLM engines that power it. The local-first claim covers the transcription layer; the assist layer is optional cloud, user-controlled.
It's desktop only today (Windows and macOS; a mobile companion is listed as coming soon), free during its current beta, and doesn't name a target industry or publish any CRM/calendar integrations — a standalone tool rather than a connected part of a sales stack. This is a genuinely different category from the AI receptionists elsewhere on this site: Conva assists a human who's already taking the call, rather than answering it instead of one.
Pricing
Free (invite-only beta) — Free through Conva's current invite-only beta — one plan, one seat, unlimited live calls. Conva's site states founding pricing will be announced before launch, and beta users will hear it first; no dollar figure is published anywhere on the site as of this check.
Verified 2026-09-03 against the vendor's pricing page. Pricing changes — always confirm with the vendor before buying.
Pros & cons
Pros
- Sub-second, speaker-labeled live transcription that keeps the rep's voice and the other party's separate (dual-channel), not one blended feed
- AI-suggested replies compose while the call is still happening and cite the specific document/passage they're drawn from, rather than an unexplained suggestion delivered after the call ends
- Local-first: call capture and transcription run on-device via an offline model — per Conva, audio never leaves the computer, and nothing is stored on its servers unless explicitly saved
- Bring-your-own API key for the cloud AI-suggest layer — a control most notetaker competitors don't offer at all
- Free during the beta; Conva says beta users will hear founding pricing before it's announced publicly
- Nothing to configure — open the app and join the call
Cons
- The AI-suggested-replies feature is opt-in cloud (bring your own LLM/speech provider API key) — the local-first privacy story applies to transcription capture, not automatically to that feature once it's enabled
- Desktop only today (Windows/macOS) — no mobile app yet, so it can't assist a call taken from a phone
- No CRM or calendar integrations published — a standalone call-assist tool rather than a connected part of a sales stack
- No named target industry or use-case specialization — generic B2B sales positioning, unlike vertically-packaged tools
- No post-beta price announced at all yet — Conva has not published what it will charge once the beta ends
- New entrant, still in invite-only beta — no long-term track record
Feature checklist
| Feature | Conva |
|---|---|
| CRM / back-office integrations | Not publishedNo integrations published as of 2026-09-03 |
| Call analytics dashboard | PartialSearchable call recordings and transcripts; no aggregate reporting/metrics dashboard published |
| Self-serve signup | ✓Free beta signup by email — nothing to configure, open the app and join a call |
| Live AI assistance while the call is happening | ✓AI-suggested replies compose while the call is still happening, not after it ends |
| On-device / local audio processing | PartialAudio capture and transcription run locally via an offline model; the AI-suggested-replies feature is a separate opt-in cloud path (bring your own API key) |
| Bring-your-own API key | ✓Bring-your-own API key for the cloud speech/LLM engines that power AI-suggested replies |
| Unlimited call length | ✓No published cap on call length, since core processing runs on-device |
| AI answers cite their source | ✓AI-suggested replies cite the specific document/passage they're drawn from |
| Dual-channel speaker separation | ✓Rep and caller kept on separate channels, not one blended transcript feed |
Frequently asked questions
- Is Conva an AI receptionist?
- No. It doesn't answer calls in place of a person — it listens alongside a sales rep who's already on the call and suggests talking points in real time. That's a different category from the AI receptionists covered elsewhere on this site.
- Does Conva store or upload call audio?
- Per Conva, capture and transcription run locally via an offline model and audio never leaves the device; nothing is stored on Conva's servers unless explicitly saved. The AI-suggested-replies feature is a separate, opt-in cloud path using your own LLM/speech provider API key.
- How much does Conva cost?
- Free during its current invite-only beta. Conva hasn't announced what it will charge after the beta ends — its site says founding pricing will be announced before launch, and beta users will hear it first.