Spokenly Review (2026): Honest Take on the Mac + iOS Dictation App with MCP

TL;DR: Spokenly is a working hybrid Mac and iOS dictation app from indie developer Vadim Akhmerov. It earns 7 / 10. The free tier runs OpenAI Whisper and NVIDIA Parakeet locally on Apple Silicon with no caps; the BYOK cloud path passes audio through whichever API key you bring (OpenAI, Deepgram, Groq, Anthropic, or Google); Spokenly Pro at $9.99 / month routes through five named subprocessors (Cerebras, Fireworks, Groq, Mistral AI, ElevenLabs) for managed cloud transcription. The MCP server is a genuine differentiator that no major dictation peer ships today. The reservations are structural: the BYOK setup tax for non-developers, the subscription-only Pro tier that breaks the category lifetime norm, the missing Windows app despite the homepage title text, the iOS keyboard reliability trade-off documented in App Store reviews, and the absence of any compliance attestation (no SOC 2, no HIPAA BAA, no ISO 27001). For developers comfortable with API-key configuration, Spokenly's free tier is one of the most generous on-device deals in 2026. For non-developers wanting no-setup polish plus lifetime pricing, Voibe at $149 lifetime is $210.64 cheaper over 3 years than Spokenly Pro with no BYOK setup and no per-token API exposure.

Disclosure: Voibe is our product. We verified Spokenly's homepage, pricing tier (Free + Pro $9.99/mo), privacy policy effective March 2, 2026, App Store listing (ID 6740315592, 4.4 / 5 from 43 ratings, v1.7.4 April 7 2026), and the best-dictation-software-2026 blog post on May 26, 2026. Where Spokenly's posture is comparable to or better than Voibe's on a specific dimension — iOS coverage, MCP server flexibility, free on-device tier — we say so.

Spokenly is a hybrid Mac and iOS dictation app built by indie developer Vadim Akhmerov (disclosed via the App Store listing). It positions itself as 'the best dictation software for Mac in 2026' on its own blog and ships three architectural modes on top of a single product:

• Local Only Mode — runs OpenAI Whisper (Large-v3) and NVIDIA Parakeet locally on Apple Silicon's Neural Engine. No network calls during transcription. Free.
• BYOK cloud mode — accepts user-supplied API keys for OpenAI, Deepgram, Groq, Anthropic, or Google. Audio passes through whichever provider you configured. Free of Spokenly fee; you pay the provider's per-call rate.
• Pro managed cloud — $9.99 / month subscription that routes audio through five named subprocessors per the privacy policy effective March 2, 2026: Cerebras, Fireworks, Groq, Mistral AI, and ElevenLabs. No BYOK required.

Three product facts to anchor the rest of this review:

• The free tier is genuinely free on-device. Whisper Large-v3 and Parakeet run locally on Apple Silicon with no time limit, no word cap, and no recurring fee. This places Spokenly's free tier alongside VoiceInk's GPL v3 build and Superwhisper's on-device modes as the most generous free dictation deals available on Apple Silicon in 2026.
• The MCP server is real. Spokenly markets itself as 'the only dictation app that runs an MCP server' — verifiable as of May 2026. AI coding agents like Claude Code, Cursor, and Codex can connect to Spokenly as a tool via the Model Context Protocol. This is a genuine wedge versus most dictation peers that integrate via accessibility APIs or workspace introspection.
• It is solo-developer indie. The privacy policy lists only admin@spokenly.app as the contact — no company name, no jurisdiction, no team page. The developer is identified on the App Store as Vadim Akhmerov. Spokenly is not VC-backed and has no public funding disclosure.

None of those facts disqualify Spokenly as a working product. They do shape who it fits and where the friction shows up. The rest of this review walks through the setup experience, accuracy testing, the MCP server scope, the iOS keyboard caveat, the privacy posture, and the pricing math — then maps each finding to a buyer profile.

The Spokenly setup experience splits cleanly along the architectural mode you choose. Local Only Mode is one-click. BYOK cloud is a five-cloud-provider exercise. Pro managed cloud is a credit-card subscription. Each path matters for different buyer profiles.

Local Only Mode Setup (Free, On-Device)

1. Download Spokenly from spokenly.app — installs as a standard Mac app.
2. Grant microphone and accessibility permissions on first launch.
3. Open settings and select “Local Only Mode” — the app downloads the Whisper Large-v3 model (~1.5GB) and Parakeet on first use.
4. Configure your global hotkey. Default is a tap-and-hold modifier; customizable to any chord.
5. Dictate into any text field, anywhere on the Mac.

Total time: about 5 minutes including model download. No accounts, no API keys, no cloud configuration. This path is the right comparison point against Voibe, VoiceInk, MacWhisper, and Superwhisper's on-device modes.

BYOK Cloud Mode Setup (Free of Spokenly Fee)

This is the setup path that creates friction for non-developers. To get BYOK cloud transcription working, you complete these steps for each provider you want to use:

1. Choose a provider. Spokenly supports OpenAI, Deepgram, Groq, Anthropic, and Google. Each has different pricing, models, and rate limits.
2. Create an account. Sign up at the provider's dashboard.
3. Add a payment method. Most providers require a credit card before issuing API keys, even for the free tier.
4. Generate an API key. Navigate to the API keys section, create a new key, copy it.
5. Paste the key into Spokenly. Settings → Cloud Providers → paste API key.
6. Pick a model. Spokenly lets you select among the provider's available speech-to-text and LLM models.
7. Configure rate limits and budget alerts. At the provider dashboard, set monthly spending caps to avoid surprises.

For developers who already manage API keys at multiple providers, this is a 10-minute exercise. For non-developers — writers, executives, students, accessibility users — this is the durable friction point. The Spokenly product treats BYOK as the headline free path, but the marginal user cost of generating five API keys at five different cloud providers is real. This is the single most actionable wedge that drives users to alternatives.

Pro Cloud Mode Setup ($9.99/month, Managed)

1. Open Spokenly, navigate to subscription settings.
2. Sign in with Apple ID for App Store billing, or via the web dashboard for direct billing.
3. Complete the $9.99/month subscription purchase.
4. Spokenly handles all cloud routing through Cerebras, Fireworks, Groq, Mistral AI, and ElevenLabs subprocessors — no per-provider configuration required.

This path eliminates the BYOK friction but introduces the subscription-only pricing model and the five-subprocessor data chain.

Spokenly Free vs Voibe Setup Side-by-Side

Voibe's setup for comparison: download, grant microphone and accessibility permissions, choose your hotkey, dictate. Total time: about 3 minutes, no model picker, no API key configuration, no cloud-provider account creation. The trade-off is that Voibe is paid ($9.90/mo or $89.10/yr or $149 lifetime) where Spokenly Local Only is free — but the setup time savings compound for users who would otherwise need BYOK cloud for cleaner formatting.

Spokenly's accuracy depends on which model you select and which mode you use. I tested across three workloads on a MacBook Air M2 (16GB) running macOS 15.3 with a Shure MV7 USB microphone — typical knowledge-worker conditions, not lab-quiet. Spokenly's own blog describes its methodology disclosure with similar setup.

1. General English Dictation (Email Drafting)

• Spokenly Local Only (Whisper Large-v3): Strong on continuous prose. Handles common acronyms (AI, HR, Q3) cleanly. Occasional capitalization drift on proper nouns. Comparable to VoiceInk, Superwhisper Standard mode, and Voibe on the same model class.
• Spokenly Local Only (Parakeet): Faster than Whisper Large-v3, with comparable accuracy on clean English audio per NVIDIA's published benchmarks. Real-time response is noticeably snappier — the use case is short-burst dictation where latency matters.
• Spokenly Pro (managed cloud): Cleanest formatting of the three — managed cloud benefits from post-processing through the LLM subprocessor chain. Minor latency penalty (~300-500ms) versus local modes due to network round-trip.

2. Technical Jargon (Code-Heavy Dictation)

• Spokenly Local Only (Whisper Large-v3): Acceptable on common programming terms (API, JSON, TypeScript) but drops accuracy on framework-specific vocabulary (Tailwind class names, Next.js routing terms). No custom-dictionary injection in the version tested.
• Spokenly + MCP server (Cursor): When connected to Cursor via MCP, the workspace context helps — Cursor's own context layer informs the transcription cleanup downstream. This is the use case the MCP server is designed for.
• Voibe Developer Mode comparison: Voibe resolves file names, folder names, and project vocabulary from the active VS Code or Cursor workspace directly into the transcription — no MCP setup required. For technical dictation, this produced fewer post-transcription edits than Spokenly Local Only without MCP.

3. Multilingual (Spanish + German Tested)

• Spokenly Local Only (Whisper Large-v3): Whisper Large-v3 is multilingual by design. Spanish dictation tested cleanly with my limited (B1-level) test corpus. German tested reasonably on common vocabulary; compound nouns and Sie-form conjugations had occasional drift. Same accuracy band as other Whisper-Large-v3-based tools.
• Spokenly Pro (managed cloud): Multilingual support via the subprocessor LLM layer — the cleanup pass helped with German compound nouns specifically.

The Honest Accuracy Verdict

Spokenly's local accuracy is in the same Whisper-Large-v3 band as every other Whisper-based Mac dictation tool — Voibe, VoiceInk, MacWhisper, Superwhisper Standard mode, Handy, Wisprtype. Accuracy differences across this peer set are usually about UI ergonomics, formatting rules, custom-vocabulary handling, and IDE integration rather than the raw transcription engine. The marketing claim of 'best accuracy' is shared across the category and means very little without an independent benchmark — none of which currently exist for Spokenly specifically.

For the typical buyer comparison, the deciding factor is not accuracy — it is what wraps the Whisper layer: setup friction, formatting controls, IDE integration, pricing, and platform reach. Spokenly wraps Whisper differently than Voibe; both work well on the same base model.

Spokenly's marketing claim of being “the only dictation app that runs an MCP server” is verifiable as of May 2026 — and it's a real product wedge that deserves an honest assessment from a competitor's review.

What MCP Is

The Model Context Protocol is an open standard from Anthropic that lets AI applications connect to data sources and tools through a uniform interface. AI coding agents like Claude Code (Anthropic), Cursor's AI features, and OpenAI's Codex all support MCP as a way to extend their capability surface — read files, query databases, call APIs, and now consume voice input.

What Spokenly's MCP Server Does

Spokenly exposes itself as an MCP tool that AI coding agents can invoke. When you're working in Claude Code or Cursor and want to dictate a prompt, the agent calls Spokenly via MCP, Spokenly transcribes the audio, and returns the text to the agent — all within the agent's context.

What the MCP Path Does That Voibe Developer Mode Doesn't

• Agent-aware integration. The transcription happens inside the agent's workflow rather than being pasted into a text field — the agent knows the voice input was dictated.
• Flexibility across agents. Any MCP-compatible agent can connect, including future agents that haven't shipped yet.
• Server-side composition. An agent can chain Spokenly with other MCP tools (file readers, search, custom workflows).

What Voibe Developer Mode Does That MCP Doesn't

• Zero configuration. Voibe Developer Mode works in Cursor and VS Code immediately without per-IDE MCP setup. The trade-off is that it's specific to those two IDEs.
• Workspace file-name resolution. Voibe reads the active workspace and resolves file names, folder names, and project-specific vocabulary directly into the transcription — “edit the auth dot tee ess file” becomes “edit the auth.ts file.” This is a transcription-layer feature that lives below the agent.
• No MCP key management. MCP servers typically require configuration files, key management, and per-agent setup. Voibe's mode is a single toggle.

Who Should Pick MCP Over Developer Mode

• Pick Spokenly MCP if you already run an MCP stack, you use multiple AI coding agents, you want to chain dictation with other MCP tools, or you're betting on MCP becoming the dominant agent-integration standard.
• Pick Voibe Developer Mode if you primarily use Cursor or VS Code, you want voice-to-code to work immediately without configuration, or you want file-name resolution to happen at the transcription layer.

Both paths solve overlapping pieces of the “voice-prompt your AI coding agent” job — they just place the integration at different points in the stack. Spokenly's MCP server is genuine engineering work and deserves credit; Voibe's Developer Mode is purpose-built for Cursor and VS Code with no configuration cost. Pick the one whose trade-off matches your workflow.

Spokenly's iOS app (App Store ID 6740315592, “Spokenly: Audio to Text AI app”) ships a custom keyboard alongside the dictation tools. App Store reviews flag recurring reliability issues with the keyboard specifically — and the developer's own reply to those reviews is the most honest acknowledgment in Spokenly's documentation.

What the App Store Reviews Say

The pattern across multiple reviews:

• Unexpected app switches when tapping the microphone button
• Recording-start failures requiring device restarts
• Inconsistent behavior across iOS versions and device generations
• Specific complaints about the local-model performance on older iPhones

What the Developer Reply Says

Vadim Akhmerov's published App Store replies (paraphrased from the public listing) acknowledge the issue and recommend a workaround: switch from local models to online (cloud) models for keyboard use. The reasoning is that local AI models stress iOS device performance more than the keyboard process can reliably handle, particularly on older devices, while cloud models offload that compute.

Why This Matters for the Buyer Decision

The developer's recommendation is honest, but it also defeats the on-device privacy benefit that draws many users to Spokenly in the first place. If reliability requires switching to online models on iOS, then the iOS keyboard is functionally a cloud product — at which point the comparison should be against other cloud iOS keyboards (Willow Voice's iOS voice keyboard, Wispr Flow's iOS keyboard) rather than against on-device alternatives. Both Willow Voice and Wispr Flow have shipped more polished iOS keyboards in 2026.

Mac vs iOS Reliability Split

The Mac app does not appear to suffer the same issues — Apple Silicon Macs have substantially more compute headroom for local models than iPhones do. The reliability concern is iOS-specific. For Mac-only users, this caveat is mostly cosmetic. For users who need polished iOS keyboard dictation, Spokenly is currently not the strongest pick.

The Voibe Position

Voibe is Mac-only by design and ships no iOS app. This is a platform-reach gap relative to Spokenly's dual Mac + iOS coverage. The trade-off: Voibe focuses entirely on Mac-native polish — no iOS keyboard reliability decisions to make, no cross-platform compromises in the local-model performance budget. For users who need iOS dictation, Spokenly or one of the polished cloud peers is the right pick; for users who do most of their work on Mac, the Mac-only design becomes a feature.

Spokenly's privacy posture splits along its three architectural modes, and the privacy policy effective March 2, 2026 is reasonably substantive — better than Wisprtype's sparse policy, less detailed than Wispr Flow's published subprocessor list. The policy's load-bearing claim is verbatim: audio recordings are “not stored” on Spokenly's servers.

Local Only Mode

• Audio captured to memory, transcribed by local Whisper Large-v3 or Parakeet, written to active text field, discarded.
• No network calls during transcription.
• This is the strongest architectural privacy posture Spokenly offers — comparable to Voibe's on-device path.

BYOK Cloud Mode

• Audio routed to whichever provider's API key you configured (OpenAI, Deepgram, Groq, Anthropic, Google).
• The data-handling posture becomes the provider's posture, not Spokenly's. Read each provider's data policy separately.
• Spokenly itself does not retain the audio per the privacy policy, but the provider may have its own retention defaults — check each provider's API terms before routing sensitive content.

Pro Managed Cloud Mode

• Audio routes through five named subprocessors: Cerebras, Fireworks, Groq, Mistral AI, and ElevenLabs.
• Each subprocessor is a separate data perimeter and contractual chain.
• The privacy policy lists these subprocessors but does not detail the contractual flow-down or which subprocessors handle which workload.

The Three Structural Caveats

1. No SOC 2, HIPAA BAA, ISO 27001, GDPR, or CCPA attestations anywhere on the privacy policy or marketing site. This is consistent with most consumer Mac dictation tools (Voibe, VoiceInk, Superwhisper, MacWhisper, Spokenly all sit at the same tier on this dimension) but disqualifies Spokenly for regulated industries.
2. No company entity or jurisdiction in the privacy policy. Only admin@spokenly.app is listed as a contact. Developer Vadim Akhmerov is identified via the App Store listing, not the privacy policy. For procurement-driven buying, this is a documentation gap relative to peers like Wispr Flow (entity disclosed) or Willow Voice (Y Combinator company page).
3. iOS keyboard reliability fix defeats on-device benefit. The developer's recommendation to switch to online models for keyboard use, while honest, changes the privacy posture for iOS-keyboard users from local-only to cloud-by-default.

For a full mode-by-mode privacy review including the Spokenly Safety Decision Tree and the audit framework, see our is Spokenly safe? investigation — sibling spoke to our is Wispr Flow safe?, is Superwhisper safe?, and is Willow Voice safe? investigations.

Spokenly's pricing is structured as a free / Pro split with no lifetime option. The Pro tier is subscription-only — a structural break from the category norm where Voibe ($149 lifetime), VoiceInk ($39.99), MacWhisper (€59), and Superwhisper ($249.99) all offer one-time pricing.

Free Tier

• Unlimited on-device Whisper Large-v3 and Parakeet transcription on Apple Silicon.
• BYOK cloud transcription via OpenAI, Deepgram, Groq, Anthropic, or Google (you supply API keys; providers bill you directly).
• MCP server access for AI coding agent integration.
• iOS keyboard (with reliability caveats noted above).

Spokenly Pro ($9.99 / month)

• Managed cloud transcription with no BYOK setup.
• Audio routes through Cerebras, Fireworks, Groq, Mistral AI, and ElevenLabs subprocessors.
• Priority support.
• No lifetime alternative. Annual billing not visibly surfaced on the pricing page as of May 2026.

3-Year Total Cost Comparison

For a full pricing-page breakdown with the BYOK cost calculator across all five named providers, see our Spokenly pricing guide.

Pros

• Genuine free on-device tier — Whisper Large-v3 and Parakeet on Apple Silicon with no caps
• Hybrid architecture spans Local Only / BYOK cloud / Pro managed cloud — different modes for different jobs
• MCP server is a real wedge no major dictation peer matches today
• Mac + iOS coverage on a single product
• Active April 2026 release cadence (v1.7.4) — not abandonware
• App Store 4.4 / 5 from 43 ratings — meaningful third-party validation
• Named developer (Vadim Akhmerov) via App Store disclosure

Cons

• BYOK setup tax is a durable friction point for non-developers
• Pro tier is subscription-only at $9.99/month — no lifetime, breaks category norm
• No SOC 2 / HIPAA BAA / ISO 27001 / GDPR / CCPA attestations
• Homepage title text says “Windows” but no Windows app ships
• iOS keyboard reliability issues documented in App Store reviews
• Pro managed cloud routes through five subprocessors — each is a separate data perimeter
• Privacy policy lists only email contact, no company entity or jurisdiction
• Solo-developer indie product — continuity risk vs commercial peers backed by funded teams

Best For

• Developers comfortable with BYOK setup and managing API keys at multiple providers
• Apple Silicon Mac users who want free unlimited on-device dictation with Whisper Large-v3 or Parakeet
• MCP power users who already run agent stacks across Claude Code, Cursor, and Codex
• iOS users willing to accept the keyboard reliability trade-off (or use online models)
• Indie-product enthusiasts who value supporting solo developers

Not Ideal For

• Non-developers who want no-setup polished dictation — Spokenly Pro $9.99/mo or Voibe $149 lifetime are the lower-friction paths
• Regulated industries (healthcare, legal, finance) requiring SOC 2 / HIPAA BAA
• Cross-platform teams needing Windows or Android
• Buyers who specifically want lifetime pricing over subscription
• Procurement-driven buying that requires documented entity, jurisdiction, and compliance attestations
• Coders who specifically use Cursor or VS Code and want zero-config voice-to-code without MCP setup — see Voibe's Developer Mode

Spokenly earns 7 / 10 as a real, working hybrid Mac and iOS dictation product. The free on-device tier is one of the most generous deals on Apple Silicon in 2026. The MCP server is genuine engineering work that no major peer matches today. The active April 2026 release cadence and named-developer transparency place it ahead of the most anonymous indie products in the category.

The reservations are structural rather than technical: BYOK setup friction, subscription-only Pro that breaks category norms, missing Windows app despite homepage title text, iOS keyboard reliability issues, and the absence of any compliance attestation. None of these issues make Spokenly unusable; together they shape the buyer profile narrowly to developers comfortable with API-key configuration on Apple Silicon Macs without regulated-industry requirements.

Who Should Pick Spokenly

• Developers who already manage API keys and run MCP agent stacks
• Mac + iOS users who value the dual-platform coverage
• Free-tier seekers on Apple Silicon who want unlimited on-device dictation
• MCP power users betting on the protocol becoming dominant

Who Should Look Elsewhere

• Non-developers wanting no-setup polish: Voibe at $9.90/mo or $89.10/yr or $149 lifetime — Mac-native, no BYOK, Developer Mode for Cursor and VS Code without MCP configuration, audited entity. Pays back versus Spokenly Pro 3-year at $210.64 in savings.
• Open-source advocates: VoiceInk at $39.99 lifetime (or free GPL v3 build from source) — closest free-on-device peer with auditable codebase.
• Power users wanting maximum customization: Superwhisper at $249.99 lifetime — Tiny / Base / Small / Standard / Parakeet local modes plus BYOK cloud, deepest customization in the category.
• Cross-platform teams needing Windows: Wispr Flow at $144/year — Mac + Windows + iOS + Android + Chrome with audited SOC 2 / ISO 27001 / HIPAA BAA stack.
• Privacy-default-protected cloud users: Willow Voice at $144/year — Private Mode is the documented (DEFAULT Opt-Out) for training, most privacy-protective cloud default among major peers.
• Regulated industries (healthcare, legal): See our best dictation software for doctors and best dictation software for lawyers guides for HIPAA-BAA paths.

The Voibe Recommendation

For Mac users comparing Spokenly and Voibe head-to-head, the choice comes down to your relationship with BYOK setup and your IDE integration needs. Spokenly's free tier is genuinely free on-device but the BYOK path adds setup friction for non-developers. Voibe's $149 lifetime is paid upfront but eliminates the BYOK question entirely — Whisper runs on-device, Developer Mode integrates with Cursor and VS Code without MCP configuration, and the pricing locks in forever. Over a 3-year horizon, Voibe is $210.64 cheaper than Spokenly Pro and trades the MCP flexibility for zero-config integration.

For the full head-to-head, see our Voibe vs Spokenly comparison. For the alternatives roundup, see our 8 best Spokenly alternatives for Mac guide.

Try Voibe for Free — install, grant microphone and accessibility permissions, dictate. No account, no credit card, no BYOK setup, no per-token API spend.