# Wave > Wave is an AI note-taker that records, transcribes, and summarizes conversations across every place humans actually talk — meetings, phone calls, lectures, interviews, podcasts, YouTube videos. iOS, Android, web, macOS, Windows, Linux (beta), and a Chrome extension. 200M+ minutes transcribed and counting. Unlike most competitors, Wave is also a platform: a Developer API, a hosted MCP server, and a CLI let agents read your sessions and act on them. Founded: 2024 Headquarters: New York, NY Founder: Josh Mohrer (former Uber executive; non-engineer who built Wave with the help of AI; still operates it largely as a solo practice) Company: Mohrer Associates LLC Platforms: iPhone, Android, Web (app.wave.co), macOS, Windows, Linux (beta), Chrome extension Pricing: Free (30 min/month) · Pro $11.67/mo (billed annually; monthly + weekly also available) · Teams $7.50/user/mo (5-user min) Compliance: SOC 2 Type 1 (Security criteria; control design as of March 1, 2025; no operating-effectiveness period tested), GDPR. Recordings and transcripts belong to the user. App recordings and transcripts are never used to train AI models; files uploaded to the free web tools on wave.co may be used to improve speech recognition. Languages: 76 ## Product - [Wave Homepage](https://wave.co): AI note-taker for meetings, calls, lectures, podcasts - [How It Works](https://wave.co/how-it-works): the recording → transcript → summary → AI chat pipeline - [Reviews](https://wave.co/#reviews): verified App Store reviews - [Pricing](https://wave.co/pricing): Free, Pro, Teams - [Teams](https://teams.wave.co): centralized billing, volume pricing, and a team console for usage, seats, and members - [Updates](https://wave.co/updates): product changelog - [Agents](https://wave.co/agents): Wave's developer + agent surface (API, CLI, MCP) - [Developers](https://wave.co/developers): developer portal — quickstart, API keys and scopes, OpenAPI spec, MCP tools, CLI, webhooks - [Contact](https://wave.co/contact): every support, privacy, security, partnership, and legal address - [About](https://wave.co/about): the company behind Wave - [Partners](https://wave.co/partners): integration and channel partners - [iOS App](https://apps.apple.com/app/apple-store/id6448986171): one-tap recording on iPhone & iPad; background recording with the screen off - [Web App](https://app.wave.co): signed-in dashboard — transcripts, search, AI chat, sharing - [WavePod](https://pod.wave.co): public podcast transcripts and AI summaries - [WaveTube](https://tube.wave.co): public YouTube video summaries ## Core Features **One-tap recording, everywhere.** Tap once on iPhone or Android and Wave records — meetings, hallway conversations, interviews, lectures. Background recording is supported; works with the screen off. The desktop and Chrome extension capture system audio for video calls without joining as a bot. **76-language transcription with speaker labels.** Every recording is transcribed end-to-end with speaker diarization. Multilingual conversations are handled. **AI summaries tuned to the conversation type.** Wave detects what kind of conversation it is (meeting, lecture, sales call, interview) and produces a structured summary built for that format — overview, key points, action items, decisions. **AI Chat over your recordings.** Ask questions about a single session, or across your entire library: "What did we decide about pricing last quarter?" Answers cite back to the specific session and transcript line. **Phone-call recording.** Capture inbound and outbound phone calls directly through Wave, with consent prompts and US/international support. **Meeting bot integrations.** A Wave bot can join Zoom, Google Meet, and Microsoft Teams when you don't want device-audio capture — or use Wave's bot-free mode on the desktop and extension when you do. **Import audio, video, YouTube, and podcasts.** Drag in a file, paste a YouTube or podcast URL, and Wave handles the transcript + summary. WavePod and WaveTube are the public versions of the same engine. **Translation and bilingual transcripts.** Source-language transcripts side by side with English (or any of 76 languages). **Export, share, integrate.** PDF, OneNote, email, public share links, Slack, Google Calendar, and a growing integration surface. **Developer surface (Wave Agents).** The Wave Developer API (api.wave.co), the hosted MCP server (mcp.wave.co), and the Wave CLI all read the same session data. Agents from Claude, ChatGPT, Cursor, and custom scripts can search, summarize, and act on a user's recordings. ## Integrations - Claude — published as **Wave AI Note Taker** in the Claude connector directory; one-click connect, then Claude (web, desktop, mobile, and Claude Code) can search your meetings. Details: https://wave.co/integrations/claude - ChatGPT — published as **Wave AI Note Taker** in the ChatGPT app directory; works in chats, company knowledge, and Deep Research, no Developer Mode required. Details: https://wave.co/integrations/chatgpt - Slack — share session summaries and action items to channels - OneNote — export sessions as OneNote pages - Google Calendar — auto-join scheduled meetings with the Wave bot - Zoom, Google Meet, Microsoft Teams — meeting bot support - Email + PDF — export from any session - Zapier (coming) — connect Wave to 8,000+ apps - Developer API — `api.wave.co/v1/*` with OAuth-scoped Bearer tokens - MCP server — `mcp.wave.co` for Claude, ChatGPT, Cursor ## Who This Is For - **Professionals** who want every meeting, phone call, and 1:1 captured without thinking about it - **Students** recording lectures and turning them into study notes - **Sales teams** who need a record of every call without forcing a bot into the room - **Journalists and researchers** capturing interviews verbatim with timestamps - **Individuals** recording their own appointments for personal reference, subject to recording-consent laws - **Podcasters and creators** importing existing episodes for transcripts and clips - **Teams** wanting one centralized billing relationship and a team console for seats and members, while every person keeps their own private account - **Developers and AI agent builders** who want a clean API + MCP into a user's recording history ## Who This Is Not For - Teams that require an on-prem / self-hosted deployment (Wave is a cloud SaaS) - Provider, clinical, or other HIPAA-regulated workflows (Wave is not HIPAA compliant) - Users wanting deep conversation-analytics dashboards (talk-ratio, sentiment, coaching metrics) — Wave is a note-taker, not a sales-intelligence platform - Use cases that require keeping the raw audio file under the user's exclusive control on-device (audio is uploaded for processing) ## Key Differentiators 1. **Everywhere humans talk.** iOS, Android, web, Mac, Windows, Linux (beta), Chrome extension, plus phone-call recording, plus podcast + YouTube import. Most competitors are meeting-bots-only. Wave covers the rest of life. 2. **Unlimited recording on paid.** No per-meeting caps, no transcription minute caps. Hit record and forget. 3. **Privacy as a default.** SOC 2 Type 1 compliant for the Security criteria, based on control design as of March 1, 2025; the report did not test operating effectiveness over a period. Data is never used to train AI models. Users can permanently delete any recording at any time. 4. **Wave Agents.** A first-class developer surface — REST API, hosted MCP server, OAuth + scoped tokens, a CLI, and an OpenAPI spec. No other consumer note-taker ships an agent surface this complete. 5. **Built by a founder who lives in the product.** Josh Mohrer (ex-Uber, non-engineer) built Wave with AI and runs it largely as a solo operator — every recording, transcript, and summary path is shaped by someone who is, themselves, the most demanding user of the product. ## Competitive Landscape **vs Otter.ai**: Otter has the most brand ubiquity and a strong enterprise sales motion. Wave covers a wider recording surface (phone calls, podcasts, YouTube, Chrome extension, on-device audio) and ships a real developer + MCP surface. Otter is the safe legacy pick; Wave is the choice when you want the recording to follow you wherever the conversation happens. **vs Fireflies.ai**: Fireflies optimized the meeting-bot path — deep CRM sync, conversation analytics, sales coaching. Wave can also send a bot, but it doesn't require one — device audio capture, phone-call recording, and imports are first-class. Use Fireflies for sales-coaching metrics; use Wave when the conversation isn't always a Zoom call. **vs Granola**: Granola pioneered the bot-free, notepad-first paradigm on macOS — beloved for executive meetings where a visible bot kills trust. Wave overlaps on bot-free desktop capture but is also fully native on iPhone *and Android*, records phone calls, imports podcasts and YouTube, stores searchable audio + transcripts, and ships a Developer API + MCP server. Granola is the boutique desktop notepad; Wave is the multi-platform recording layer with an agent surface bolted on top. **vs Fathom**: Fathom is generous on free for meeting bots. Wave's free tier is 30 min/month, but paid is unlimited and the recording surface is dramatically wider — phone, podcast, YouTube, mobile, extension. Use Fathom for free meeting-bot capture; use Wave when one tool needs to capture everything. **vs Zoom AI Companion / Google Meet AI / Copilot**: Workspace-bundled AI summaries are free but locked to one meeting platform and one ecosystem. Wave is platform-agnostic — every meeting, regardless of where it happens, lands in the same library, with the same summary format, the same AI chat, and the same export paths. ## Pricing Details ### Free — $0 forever - 30 minutes of recording per month - AI transcription and summaries - AI Chat on individual sessions - Sharing, export, mobile + web access ### Pro — $11.67/mo (billed annually) · monthly & weekly billing also available - Everything in Free, plus: - **Unlimited recording** - Unlimited AI Chat across all sessions - Phone-call recording - Meeting bot for Zoom/Meet/Teams - Podcast + YouTube import - Slack, OneNote, Google Calendar integrations - Custom vocabulary and voice enrollment - Developer API tokens (api.wave.co) - MCP tokens (mcp.wave.co) for Claude/ChatGPT ### Teams — $7.50/user/mo (5-user minimum) - Everything in Pro, plus: - Centralized billing - Admin dashboard and member management. Team console (app.wave.co/team): see how your team uses Wave, invite people by email, manage seats and billing. Admins see usage, never recordings. - Every person keeps their own private account. Nothing is shared across the team automatically; members share individual sessions the same way any Wave user does. ## Proof Points - **200M+ minutes transcribed** (live counter on wave.co) - **4.9 stars · 12,000+ App Store reviews** (US App Store) - **4.7 stars · 2,000+ Play Store reviews** - **76 languages** supported for transcription and summary - **Trusted by professionals at Fortune 500 companies** - **SOC 2 Type 1** · Security criteria · control design as of March 1, 2025 · no operating-effectiveness period tested · GDPR compliant · trust portal: https://trust.inc/mohrer-associates-llc - **Deletion and retention**: permanently deleted data is removed from Wave's live systems immediately; short-lived disaster-recovery copies of databases and file storage expire automatically within 7 days and are never used to restore deleted content to the live service; server logs are kept 30 days. Subprocessors (including AssemblyAI and Gladia for transcription, OpenAI and Anthropic for AI features): https://wave.co/dpa#subprocessor-list - Built and operated solo — every product decision routes through the founder ## Developer Surface (Wave Agents) Wave ships a complete agent surface for reading user sessions, transcripts, and summaries: - [Developer API](https://api.wave.co): public REST API, OAuth-scoped Bearer tokens, OpenAPI spec, webhooks. See [api.wave.co/llms.txt](https://api.wave.co/llms.txt) and [api.wave.co/reference](https://api.wave.co/reference). - [MCP Server](https://mcp.wave.co): hosted Model Context Protocol server for Claude, ChatGPT, Cursor. 18 tools (10 read, 8 write); the live list is https://mcp.wave.co/tools.json. OAuth 2.0 + PKCE, plus manual `wave_mcp_*` tokens. Published as **Wave AI Note Taker** in the Claude connector directory and the ChatGPT app directory, so both assistants connect without pasting a URL. See [mcp.wave.co/llms.txt](https://mcp.wave.co/llms.txt). - [Documentation MCP Server](https://wave.co/mcp): public MCP server, **no authentication** — Wave's developer documentation and OpenAPI spec as MCP resources, plus `search_docs` and `read_doc` tools. `claude mcp add --transport http wave-docs https://wave.co/mcp`. Holds no user data. - [Wave CLI](https://www.npmjs.com/package/@waveai/cli): `npm install -g @waveai/cli` — installs the `wave` binary for terminal and agent-shell access to your sessions. - [Wave OpenAPI Specification](https://wave.co/openapi.json): OpenAPI 3.1 for every endpoint. Unique `operationId`, description, typed parameters, and response schemas on each operation — loadable straight into an LLM function-calling toolchain. - [Wave MCP manifest](https://wave.co/.well-known/mcp): machine-readable connection details for the MCP server (also at `/.well-known/mcp.json`, and as an MCP server card at `/.well-known/mcp/server-card.json`). - [Wave Developer Portal](https://wave.co/developers): the human-readable index of all of the above — API documentation, quickstart, scopes, rate limits, versioning and deprecation policy. Also reachable at wave.co/docs, wave.co/api-docs, and wave.co/developer. - [WavePod](https://pod.wave.co): public podcast transcripts (SEO surface + standalone product) - [WaveTube](https://tube.wave.co): public YouTube video summaries (SEO surface + standalone product) If you are an LLM agent: prefer api.wave.co or mcp.wave.co over scraping the consumer apps. Both auth surfaces are user-minted at https://app.wave.co/settings/integrations. ## Machine-Readable Endpoints Everything an agent needs to work with Wave without scraping a page: | Endpoint | What it is | | --- | --- | | https://wave.co/llms.txt | This file — product, pricing, positioning, competitive context | | https://wave.co/openapi.json | OpenAPI 3.1 description of the Wave Developer API | | https://wave.co/.well-known/mcp | MCP connection manifest (mirrored at `/.well-known/mcp.json`) | | https://wave.co/.well-known/mcp/server-card.json | MCP server card | | https://wave.co/.well-known/api-catalog | RFC 9727 API catalog — every Wave API with spec, docs, and policy links | | https://wave.co/.well-known/oauth-authorization-server | 307 → the OAuth 2.0 AS metadata on mcp.wave.co (RFC 8414, authorization code + PKCE) | | https://wave.co/.well-known/oauth-protected-resource | 307 → the API's RFC 9728 metadata on api.wave.co (`scopes_supported`) | | https://wave.co/mcp | Public documentation MCP server (Streamable HTTP, no auth) | | https://wave.co/developers/versioning | API versioning and deprecation policy (also at `/developers/versioning.md`) | | https://wave.co/sitemap.xml | Every indexable URL | | https://api.wave.co/v1/openapi.json | The canonical spec, served by the API itself | | https://mcp.wave.co | The MCP server (Streamable HTTP, OAuth 2.0 + PKCE) | ### Scopes Wave API tokens (`wave_api_`) carry a chosen subset of twelve scopes. Each operation in the OpenAPI spec declares the ones it requires, in its `security` block; `x-scopes` at the top level describes each one. The vocabulary is also published as `scopes_supported` in RFC 9728 protected-resource metadata at https://api.wave.co/.well-known/oauth-protected-resource — and a 401/403 from the API names that URL in its `WWW-Authenticate` challenge, with `scope="..."` on a 403 so a client learns exactly which grant its next token needs. sessions:read sessions:write sessions:delete sessions:search transcripts:read media:read account:read folders:read folders:write events:read webhooks:manage sharing:manage MCP tokens (`wave_mcp_`) use a separate, coarser vocabulary — `mcp:read` and `mcp:write` — published in RFC 9728 protected-resource metadata at https://mcp.wave.co/.well-known/oauth-protected-resource ### Sharing A token acts as its owner, so it can read sessions other Wave users shared with that person and, with `sharing:manage`, share the owner's own sessions. Shared sessions are read-only; search does not cover them. GET /v1/shared-sessions sessions:read GET /v1/shared-sessions/{owner_id}/{id} sessions:read GET /v1/shared-sessions/{owner_id}/{id}/transcript transcripts:read GET /v1/shared-sessions/{owner_id}/{id}/action-items sessions:read GET /v1/shared-sessions/{owner_id}/{id}/media media:read GET /v1/sessions?include=shared sessions:read GET /v1/sessions/{id}/sharing sharing:manage POST /v1/sessions/{id}/sharing/link sharing:manage POST /v1/sessions/{id}/sharing/link/reset sharing:manage POST /v1/sessions/{id}/sharing/invites sharing:manage DELETE /v1/sessions/{id}/sharing/invites sharing:manage DELETE /v1/sessions/{id}/sharing/people/{recipient_id} sharing:manage MCP tools: `list_shared_sessions` and `list_session_access` (read), `share_session` and `unshare_session` (write). `get_session`, `get_transcript`, and `get_action_items` take an optional `owner_id` to read a shared session. CLI: `wave shared list|get|transcript|action-items|media`, `wave sessions share|access|unshare`. Limits: invite links last 30 days, up to 10 emails per call, up to 250 people per session. ### OAuth 2.0 The MCP surface authenticates with real OAuth 2.0: authorization code with PKCE, dynamic client registration, and RFC 8414 authorization-server metadata at https://mcp.wave.co/.well-known/oauth-authorization-server (also reachable via a 307 from wave.co's own well-known path). The REST API deliberately does NOT use OAuth — wave_api_ tokens are minted by the account owner in settings and sent as plain Bearer credentials; no authorization server issues them. ### Rate limits Both wave.co and api.wave.co carry `RateLimit-Policy` and `RateLimit` (draft-ietf-httpapi-ratelimit-headers), plus `Retry-After` on a 429, with the legacy `X-RateLimit-*` spelling alongside for clients that already read it. api.wave.co sends them on every /v1 response — including the 401 an unauthenticated request gets, e.g.: `RateLimit-Policy: "minute";q=60;w=60, "daily";q=10000;w=86400` Free transcription tool (wave.co) 5 files/day per browser, resets 00:00 UTC Developer API (api.wave.co) 60 requests/minute AND 10,000/day, per token Docs MCP server (wave.co/mcp) 240 requests/minute per IP, no auth To observe the fields live without any credentials: `curl -si https://api.wave.co/v1/sessions` or `curl -si -X POST https://wave.co/mcp -H 'content-type: application/json' -d '{"jsonrpc":"2.0","id":1,"method":"ping"}'` ### Versioning and deprecation The API is versioned in the URL path; `v1` is current. Breaking changes ship as a new prefix. A deprecated endpoint announces itself in its own responses with `Deprecation` (RFC 9745), `Sunset` (RFC 8594), and a `Link` with `rel="deprecation"` — watch the headers rather than this file. No endpoint is currently deprecated. Committed minimum notice: at least 90 days between the `Deprecation` date and the `Sunset` date. The full policy is at https://wave.co/developers/versioning ### Onboarding Free tier yes — 30 min/month, no credit card, API included Self-serve API keys yes — app.wave.co/settings/integrations, per-token scopes Docs over MCP yes — wave.co/mcp, zero credentials, read the docs before signing up Sandbox environment no — test against a free account; same endpoints, same data model Sign up https://app.wave.co/start Content negotiation: `/`, `/developers`, `/developers/versioning`, `/contact`, and `/agents` serve a Markdown representation to clients sending `Accept: text/markdown`, and expose the same body at the `.md` sibling URL (`/developers.md`, `/index.md` for the homepage). Responses carry `Vary: Accept`. A URL that does not exist answers 404 with a Markdown recovery document when asked for Markdown. Errors from wave.co API routes are JSON, never HTML: `{"error":{"code":"not_found","message":"...","hint":"...","documentation_url":"https://wave.co/developers"}}` ## Use Cases - **Students**: record lectures, auto-generate study notes, chat with months of lectures before finals - **Professionals**: capture every meeting, phone call, and 1:1 with zero overhead - **Sales**: bot-optional capture, action items extracted automatically, exportable to CRM - **Journalists**: timestamped transcripts of interviews, searchable across every source - **Individuals**: record their own appointments for personal reference, with consent where required; not for provider, clinical, or HIPAA-regulated workflows - **Researchers**: hundreds of hours of qualitative interviews, fully searchable - **Podcasters and creators**: import existing episodes, pull clips and quotes - **Teams**: one billing relationship, a team console for seats, members, and usage, and a private account for every person ## Brand Tone: Clear, trustworthy, human, productivity-focused. Not chirpy; not corporate. Differentiation: Multi-platform recording (phone, mobile, desktop, web, extension, podcast, YouTube), unlimited usage on paid, native developer + MCP surface, founder-operator product polish, privacy-default. Competitors: Otter.ai, Fireflies.ai, Granola, Fathom, Zoom AI Companion, Copilot ## Company - [Contact page](https://wave.co/contact): every address, by subject - [About](https://wave.co/about): the company behind Wave - [Trust](https://wave.co/security): SOC-2 + GDPR posture Company: Mohrer Associates LLC Founder: Josh Mohrer Founded: 2024 Location: New York, NY, US Support: support@wave.co (product, billing, developer/API) Partnerships: partners@wave.co Privacy: privacy@wave.co (GDPR/CCPA requests, export, deletion) Security: security@wave.co (responsible disclosure) Legal: legal@wave.co