Market category: MCP Servers
AssemblyAI provides voice AI infrastructure that helps developers build applications for processing and understanding spoken conversations.
Overview
AssemblyAI provides voice AI infrastructure for developers and organizations building voice-enabled applications. Its capabilities help teams convert speech into text, derive structured insights from conversations, and create natural spoken interactions, with tools for deployment, security, and compliance.
Mission statement
Help developers build voice-enabled products by making voice AI capabilities accessible, reliable, and scalable.
One of 19 AI agents in AI InfrastructureOne of 83 AI agents in MCP Servers
Work at AssemblyAI? Decide what AI assistants read about you.
Claim AssemblyAIUse your email at assemblyai.com.What we offer
Pre-recorded Speech-to-Text API
Turn recorded speech into accurate, customizable transcripts and structured insights for applications and workflows.
assemblyai.com →Realtime Speech-to-Text API
Turn live speech into fast, accurate transcripts for responsive voice applications, captions, and real-time assistance.
assemblyai.com →Sync Speech-to-Text API
Get accurate transcripts from short audio clips in one fast request, without polling or managing a transcription session.
assemblyai.com →Dictation API
Turn speech into accurate, ready-to-use text quickly, with customized formatting and a verbatim transcript included.
assemblyai.com →Speech Understanding API
Turn conversations into structured insights that help teams analyze, organize, and act on audio data.
assemblyai.com →Guardrails API
Protect voice data and keep transcription outputs clean with built-in privacy, safety, and operational controls.
assemblyai.com →LLM Gateway
Apply and compare language models on voice data and transcripts through one API, with model choice, provider rates, and task-specific options.
assemblyai.com →Text-to-Speech API
Create natural, fast voice-agent responses that keep callers engaged and scale with call volume.
assemblyai.com →Voice AI Cloud
Build and scale production Voice AI applications through a managed API, without managing GPUs, infrastructure, or capacity planning.
assemblyai.com →Self-Hosted Voice AI
Run AssemblyAI speech-to-text in your own environment to keep audio under your control, meet residency requirements, and reduce network latency.
assemblyai.com →Blurt
Dictate naturally and get cleaned-up text inserted into the app you’re using, with multiple languages and writing styles supported.
assemblyai.com →AssemblyAI MCP Server
Transcribe, summarize, and analyze audio from compatible AI clients using an AssemblyAI account, without writing code for the task.
assemblyai.com →Docs MCP Server
Helps coding agents find current AssemblyAI documentation and write more accurate integration code.
assemblyai.com →AssemblyAI Playground
Test AssemblyAI Voice AI models on audio without code and explore transcription, speaker identification, and speech understanding features.
assemblyai.com →More information about our offering
Pre-recorded Speech-to-Text API
AssemblyAI’s Pre-recorded Speech-to-Text API converts recorded audio into clean, customizable transcripts and structured insights. Choose Universal-3.5 Pro for complex, domain-specific audio or Universal-2 for high-volume transcription across 99 languages. Add speaker labeling, language detection, prompting, medical transcription, summarization, sentiment analysis, entity detection, and PII redaction to requests.
- Transcribe Recorded Audio AccuratelySelect a model suited to the audio and workload, from complex domain-specific recordings to high-volume transcription.
- Transcribe Audio Across LanguagesBuild multilingual transcription workflows using one API, with language coverage and code-switching support depending on the model.
- Attribute Transcript Segments To SpeakersIdentify who said each utterance and when, making recorded conversations easier to review and analyze.
- Detect Languages AutomaticallyAvoid manual language selection and routing, with controls for expected languages and fallback behavior.
- Guide Transcription With ContextSupply contextual information to help the model handle terms and subject matter relevant to a recording.
- Improve Recognition Of Key TermsEmphasize relevant terminology in the request to support recognition of names and specialized vocabulary.
- Transcribe Medical Audio More AccuratelyImprove recognition of medical terminology in healthcare conversations and clinical recordings.
- Recognize Clinical TerminologyImprove transcription of specialized medical language in clinical recordings.
- Extract Timestamped EntitiesConvert recognized entities into typed fields with timestamps for downstream use.
- Redact Personal InformationReduce exposure of sensitive information by removing it from transcript text and, where enabled, bleeping it from audio.
- Identify Speakers By Name Or RoleAssociate speakers with names or roles using audio context.
- Summarize Recordings And Create ChaptersTurn long recordings into concise summaries and timestamped sections that help users navigate the content.
- Extract Timestamped Action ItemsIdentify follow-ups and return them with source quotes and timestamps.
- Classify Sentiment By SentenceReview how sentiment varies throughout a transcript, including by speaker when diarization is enabled.
- Detect And Moderate Sensitive ContentIdentify sensitive content and clean up offensive language before transcript content is used downstream.
- Classify Topics In RecordingsOrganize recorded audio by classifying its subject matter against standard IAB Content Taxonomy categories.
- Find Relevant Phrases With TimestampsIdentify and rank important phrases, with timestamps to locate them in the audio.
- Standardize Transcript FormattingNormalize common data formats so dates, phone numbers, currency, and URLs are ready for downstream systems.
- Translate Transcripts Into Other LanguagesTranslate transcript text as part of the request and choose a formal or informal style.
Realtime Speech-to-Text API
A streaming API that transcribes live audio and video into partial and final text for voice agents, live captions, notetakers, agent assist, and other real-time applications. Its models offer low-latency transcription. Universal-3.6 Pro Realtime supports 32 languages and features including turn detection, speaker labels, context carryover, and keyterm prompting.
- Get Live Transcripts FastReceive text while a speaker is still talking, enabling responsive voice agents, live captions, and other applications that depend on timely transcripts.
- Detect When Speakers FinishIdentify the end of a spoken turn so applications can respond promptly without relying only on pauses.
- Transcribe Across 32 LanguagesSupport multilingual voice experiences with real-time models covering up to 32 languages and automatic language detection on supported models.
- Carry Context Across ConversationsRetain conversational context across an ongoing session to support more consistent realtime transcription.
- Boost Recognition Of Important TermsGuide supported models toward important names, terminology, and context so transcripts better fit the application or domain.
- Identify Speakers In Live AudioSeparate spoken turns by speaker and label who said what during live conversations.
- Connect To Existing Voice StacksIntegrate realtime transcription into supported orchestration frameworks and voice application stacks.
- Scale With Unlimited StreamsRun concurrent transcription streams without listed concurrency caps.
- Isolate Voices In AudioUse the supported model’s voice isolation capability to focus realtime transcription on speech.
- Format Transcripts AutomaticallyReturn readable live transcripts with automatic punctuation, casing, and supported formatting.
- Handle Language SwitchingTranscribe supported multilingual conversations that switch between languages.
- Recognize Medical TerminologyAdd Medical Mode to supported realtime models for clinical terminology recognition and noise-resilient transcription.
Sync Speech-to-Text API
A synchronous speech-to-text API for short audio clips. Send audio in one HTTP request and receive a finished transcript in the same response, typically in about 134 ms at P50. Built on Universal-3.5 Pro, it supports clips up to 20 minutes and 40 MB, with per-word text, timing, and confidence in every response. Keyterms and conversation context are supported.
- Get Transcripts In One RequestComplete transcription in a single request-response call, avoiding job management and polling.
- Receive Finished Text In ~134 MsReturn a finished transcript quickly for dictation, push-to-talk commands, voicemail, and voice agents with their own turn detection.
- Use Flagship Transcription AccuracyGet the stated Universal-3.5 Pro accuracy with a synchronous, single-request endpoint.
- Transcribe Clips Up To 20 MinutesHandle bounded audio clips in one request, including recordings up to 20 minutes.
- Use Word-Level Timing And ConfidenceAccess timing and confidence alongside each word for applications that use detailed transcript information.
- Transcribe In 18 Supported LanguagesBuild voice input for users of supported languages, including English, Spanish, French, German, Japanese, Mandarin, Hindi, and Turkish.
- Guide Transcription With KeytermsUse keyterms and conversation context to provide relevant information for transcription.
Dictation API
The Dictation API turns a spoken utterance into finished, formatted text in a single request. It combines transcription with an LLM cleanup pass, returning the verbatim transcript alongside cleaned-up text. Developers can customize transcription with context and key terms, and shape rewritten output with formatting instructions. It supports 32 languages with automatic language detection and code-switching, and returns finished text in under a second.
- Get Finished Text In One CallSend a spoken utterance and receive formatted, cleaned-up text in the same request.
- Clean Up Speech AutomaticallyReturn text with filler removed and punctuation and capitalization applied, while retaining the speaker’s phrasing and tone.
- Access Both Text VersionsUse the original words and finished rewrite from the same response, with the verbatim transcript separate from the rewrite.
- Format Text To Match Your WorkflowSpecify the structure or style an application needs, such as a clinical note.
- Receive Text In Under A SecondGet finished text quickly after a spoken utterance, supporting responsive voice input.
- Guide Recognition With Context And TermsProvide contextual guidance and important terms to help transcribe vocabulary relevant to a product or workflow.
- Dictate Across 32 LanguagesLet users dictate in supported languages, including mixed-language speech, without manual language selection.
- Transcribe With Universal-3.5 ProUse the Universal-3.5 Pro speech model for short-form dictation transcription.
Speech Understanding API
Turn raw audio into structured intelligence with purpose-built speech models in a single API call. Combine transcription with features that identify speakers, extract entities and action items, analyze sentiment and topics, summarize recordings, translate transcripts, and redact sensitive information. Choose the features needed without building separate processing pipelines.
- Combine Features In One RequestAdd multiple speech understanding capabilities to the same transcription request without building separate processing pipelines.
- Separate Speakers In TranscriptsOrganize conversation audio into speaker-labeled segments with timestamps.
- Summarize And Organize RecordingsCreate concise summaries and timestamped sections that help people navigate and review recordings.
- Extract Typed Entities With TimestampsGet structured entity data with categories and precise timestamps for downstream applications.
- Redact Personal Information From AudioRemove personally identifiable information from transcript text and bleep it from audio in one API call.
- Identify Speakers By NameAttribute conversation segments to names or roles instead of generic speaker labels.
- Extract Timestamped Follow-UpsSurface follow-ups with timestamps and supporting quotes.
- Analyze Sentiment By SentenceAdd sentence-level sentiment and, when speaker labels are enabled, connect results to individual speakers.
- Detect And Transcribe 99 LanguagesIdentify the dominant language and transcribe without manual selection, with options to restrict detection and set a fallback.
- Translate Transcripts Into 86 LanguagesProduce translated transcripts in the same request and select a formal or informal style.
- Surface Relevant PhrasesFind important concepts and phrases in recordings without manually searching raw transcripts.
- Classify Audio TopicsCategorize audio against standard IAB Content Taxonomy categories to organize and analyze recordings.
- Normalize Transcript FormattingConvert dates, phone numbers, currency, and URLs into standardized formats.
- Detect Sensitive And Offensive ContentIdentify sensitive content and clean up offensive language automatically.
Guardrails API
Guardrails provides safety, privacy, and compliance controls at the transcription layer through one API. It can redact personally identifiable information from transcript text and audio, moderate sensitive content, filter profanity, and apply speech threshold and transcript start/end controls. It supports GDPR and PCI programs, with a BAA available for customers processing PHI.
- Remove Sensitive Data From TranscriptsRedact personal information in transcript text within the transcription request, helping reduce sensitive data in downstream systems and support privacy and compliance workflows.
- Bleep Personal Information In AudioCreate audio with sensitive spans bleeped in the same request, supporting safer storage, sharing, and compliance archiving.
- Flag Sensitive Content Against Your PolicyIdentify potentially sensitive material and review flagged text, timestamps, confidence, and severity to support content review and policy enforcement.
- Clean Profanity From TranscriptsAutomatically filter offensive language from transcription output to help keep content clean.
- Transcribe Only Files With Enough SpeechUse a minimum spoken-audio threshold to validate whether an input should be transcribed.
- Limit Transcription To Spoken SectionsSet transcript start and end controls to focus on sections containing speech.
LLM Gateway
LLM Gateway provides unified access to open and frontier language models through one API for voice data and transcripts. Use more than 50 models from multiple providers for dictation, information extraction, summarization, and text correction. Requests use an OpenAI-compatible interface, and the Gateway shares an API key with AssemblyAI Speech-to-Text. Compare model context, pricing, capabilities, tasks, providers, and regions, and switch models by changing a model identifier.
- Access Multiple Models Through One APIUse one interface to apply different language models to voice data and transcripts across supported providers.
- Choose Models For Your WorkflowsCompare model options and select based on published capabilities, task recommendations, pricing, provider, and available region.
- Integrate With An OpenAI-Compatible APIConnect existing SDK workflows and switch the selected model by changing its identifier.
- Process Transcripts For Different TasksChoose models for transcript workflows such as cleanup, extraction, summarization, correction, generation, and formatting.
- Use One Key For Speech And LLMsUse the same API key for Speech-to-Text and LLM Gateway requests.
- Enable Automatic Model FallbacksUse the Gateway’s automatic fallback capability when working with supported models.
- Pay Provider Rates With No MarkupPay selected model provider rates without Gateway markup, based on input and output tokens and with no minimum commitment.
- Compare Model Details Before ChoosingReview model context, costs, capabilities, supported tasks, providers, and regions to inform selection.
- Use Zero Retention For Eligible ModelsFor Anthropic and Google models under a signed BAA, the Gateway offers zero data retention.
- Select Models By Available RegionCompare listed regions and choose available US, EU, or global options where supported.
Text-to-Speech API
AssemblyAI-TTS-1 converts text into natural-sounding speech for voice agents and live calls. Trained on millions of hours of audio, including real conversations, it streams audio from the first byte, supports six languages, and offers a voice library and instant voice cloning. The API is designed for low-latency responses and concurrency that scales with call volume. Use it with LiveKit or Pipecat, or build an integration over HTTP or WebSocket.
- Create More Natural Voice-Agent ConversationsGenerate speech designed for real conversations, helping voice agents sound more natural on live calls.
- Start Speaking From The First ByteBegin streaming audio quickly so voice agents can respond sooner during live calls.
- Scale Calls Without Concurrency LimitsScale concurrency with call volume to accommodate call spikes without negotiating a commitment.
- Clone A Voice From A 30-Second ClipCreate a branded agent voice from a short audio clip, with permission confirmed at upload.
- Speak Across Six Supported LanguagesBuild voice agents that speak six supported languages with the appropriate accent for the selected language.
- Choose From Thousands Of VoicesExplore and audition voices across languages, accents, and regions.
- Integrate Through Frameworks Or APIsConnect through LiveKit or Pipecat, or use HTTP or WebSocket for a custom integration.
Voice AI Cloud
Voice AI Cloud is AssemblyAI’s managed inference platform for production Voice AI applications. It provides access to AssemblyAI models through an API and SDKs, with infrastructure that scales automatically from zero to thousands of concurrent streams. AssemblyAI handles model serving, GPU provisioning, capacity planning, reliability, and infrastructure updates. The service offers usage-based, per-second pricing with no minimums or long-term contracts, a 99.9% uptime SLA, multi-region redundancy, automatic failover, and SOC 2 and GDPR support.
- Run Models Without Managing InfrastructureUse managed inference instead of building and operating model-serving infrastructure or provisioning GPUs.
- Scale From Zero To Thousands Of StreamsAdjust capacity automatically as voice traffic changes, supporting production workloads and spikes without capacity planning or GPU provisioning.
- Maintain Resilient Production WorkloadsRely on the stated uptime SLA, multi-region redundancy, and automatic failover for production applications.
- Build Voice Apps Through One APIIntegrate managed inference and AssemblyAI models through a single API and available SDKs.
- Pay Only For UsageUse per-second pricing without minimum commitments or long-term contracts, paying for usage rather than over-provisioned capacity.
- Support Enterprise Compliance RequirementsUse cloud infrastructure with SOC 2 and GDPR support without additional customer-side setup.
- Skip Infrastructure MaintenanceAvoid managing capacity planning and infrastructure updates; AssemblyAI handles scaling, reliability, and infrastructure optimizations.
Self-Hosted Voice AI
Run AssemblyAI speech-to-text models within your own infrastructure, including a VPC, on-premises environment, or supported container orchestration platform. Keep audio inside your controlled perimeter while managing deployment and scaling to suit your traffic. Self-hosted deployments provide the Realtime Speech-to-Text API with Universal-3.5 Pro Realtime and the Sync Speech-to-Text API powered by Universal-3.5 Pro. Pricing is usage-based, matching the cloud service with no self-hosting premium.
- Run Models In Your EnvironmentDeploy AssemblyAI models within your infrastructure, keeping the voice processing stack close to your applications and under your operational control.
- Keep Audio Within Your PerimeterProcess audio inside your controlled environment to retain data sovereignty and address regulatory and data residency requirements.
- Transcribe Live Streams And Short ClipsChoose realtime streaming for live audio or the Sync API for short clips, with both running inside your environment.
- Reduce Latency Near Your TrafficRun processing closer to users and infrastructure; the site reports potential latency reductions of 50–200 ms in regions far from cloud endpoints.
- Tune Scaling To Your TrafficAdjust deployment scaling to fit traffic patterns and use provided metrics and observability to support autoscaling.
- Deploy On Your Orchestration PlatformUse Kubernetes, AWS ECS, or another existing platform, with a Docker Compose demo and engineering assistance for production deployments.
- Pay For Active SessionsUse cloud-equivalent usage-based pricing without a self-hosting premium, with daily billing and volume discounts available.
- Choose A Supported GPU ConfigurationSelect from supported GPU configurations; the deployment page states each instance handles up to 48 concurrent streams without runtime degradation.
- Apply Usage To AWS CommitmentsRun an AssemblyAI agreement through AWS Marketplace so eligible usage counts toward AWS committed spend.
Blurt
Blurt is a free, open-source dictation app for macOS 14 and later. Hold a keyboard shortcut, speak, and Blurt inserts text into the focused app. It uses AssemblyAI’s Dictation API to transcribe and clean up text, removing filler words and resolving self-corrections. Users can choose or define writing styles, dictate in 19 languages, and code-switch mid-sentence. Blurt is MIT licensed; users provide their own AssemblyAI API key, and transcription usage is billed separately.
- Dictate Into Any Focused AppCapture speech with a hotkey and have text appear in the active text field across the apps you already use.
- Turn Speech Into Clean TextSpeak naturally, including filler words or corrections, and get cleaned-up text ready to use.
- Match Text To Your StyleChoose or define a writing style for the context, such as a casual message or formal email.
- Get Dictation Back QuicklyFast transcription and cleanup help text appear promptly after speaking, without waiting on a job queue.
- Capture Speech More AccuratelyDesigned to handle accents, code, and domain jargon; the product page claims up to 30% fewer hallucinations than Whisper.
- Fork And Customize The AppFork and modify the MIT-licensed source. Blurt is free to use, but users provide their own API key and pay for transcription usage.
- Dictate Across 19 LanguagesSpeak in supported languages and switch mid-sentence; Blurt follows language changes without a setting change.
- Spell Specialized Terms CorrectlyAdd product names, acronyms, and function names to custom vocabulary to help them come out correctly.
AssemblyAI MCP Server
AssemblyAI MCP Server is a remote Streamable HTTP server that connects MCP-compatible clients to a user’s AssemblyAI account. Clients can submit and retrieve transcripts, summarize and analyze audio, and delete transcripts using AssemblyAI tools. Connect at https://mcp.assemblyai.com/mcp and authenticate with AssemblyAI OAuth sign-in or an API key. Usage is billed at normal API rates.
Category: MCP Servers
- Connect AI Clients To AssemblyAIUse supported MCP clients to access AssemblyAI audio transcription and analysis tools through the remote server.
- Transcribe Audio From An MCP ClientSubmit audio for transcription through an MCP client connected to the user’s AssemblyAI account.
- Authenticate With OAuth Or An API KeyChoose OAuth sign-in or API-key authorization to connect the MCP client to the user’s account.
- Summarize Transcripts In Your AccountRequest transcript summaries through a compatible MCP client using audio processed on the user’s account.
- Analyze Transcripts Through MCPRun transcript analysis through an MCP client using the understand_transcript tool.
- Retrieve Transcripts Through MCPGet transcript results from the user’s account through a compatible MCP client.
- Delete Transcripts From Your AccountUse the delete_transcript tool to remove transcripts from the user’s account.
- Pay At Normal API RatesMCP server usage follows AssemblyAI’s normal API-rate billing.
Docs MCP Server
A free, unauthenticated MCP server that gives coding agents access to AssemblyAI’s live documentation. It searches and reads the docs to help agents such as Claude Code and Cursor write correct AssemblyAI code.
Category: MCP Servers
- Find And Read Current DocumentationLets coding agents search and read live documentation to use relevant information when generating code.
- Connect Coding Agents To AssemblyAI DocsProvides coding agents access to AssemblyAI documentation through MCP, supporting more accurate integration code.
- Access Docs Without Cost Or AuthenticationCoding agents can use the documentation server without a fee or authentication.
AssemblyAI Playground
AssemblyAI Playground is a no-code environment for testing Voice AI models on audio. Choose a sample, upload an audio file, or provide a URL, then configure transcription and speech understanding options. The playground supports model selection, language detection, prompts, key terms, speaker settings, and analysis features including translation, topic detection, moderation, PII redaction, sentiment analysis, summarization, and action items.
- Test Voice AI Models Without CodeTry AssemblyAI Voice AI models in a browser-based playground without writing code.
- Configure Models And TranscriptionAdjust model and transcription settings, including prompts, key terms, and specialized modes, to test audio with different contexts.
- Test Audio From Multiple SourcesChoose a supplied sample, upload audio, or enter a URL to explore model behavior across inputs and scenarios.
- Test Transcription Across 18 LanguagesExplore transcription in 18 languages, including native code-switching, and use automatic language detection.
- Explore Speaker And Channel SettingsTest speaker-related options, including diarization, labels, identification, multichannel audio, and automatic detection.
- Summarize AudioExplore the summarization option to turn audio into a summary in the playground.
- Try PII RedactionTest PII redaction as part of the playground’s speech understanding features.
- Explore Audio SentimentTry sentiment analysis on audio using the playground’s speech understanding options.
- Try Audio TranslationExplore translation through the playground’s speech understanding controls.
- Explore Audio TopicsTest topic detection as part of the playground’s speech understanding options.
- Test Content ModerationExplore content moderation through the playground’s speech understanding controls.
- Test Profanity FilteringExplore profanity filtering through the playground’s speech understanding controls.
- Identify Action ItemsTry the action items option among the playground’s speech understanding features.
Sources
Methodology and sourcing behind the figures and links shown above.
This is a public preview. Whoever claims it decides what it shows.
This page was built from public information. Claim it and the AI agent behind it learns far more than this page says; that stays in your workspace, is never shown to visitors or to AI assistants, and nothing here changes without your approval.
Own this company? You choose what is listed here: the summary and offers, which comparisons appear, the FAQ, or whether the page is listed at all. Unlisting takes one switch.
Built from public web sources. Request removal →
How AI sees this company
This is what AI systems and crawlers receive for this page — the metadata and structured data, and the Markdown version, served alongside the human-readable content.