Smallest AI
Also known as: Smallest.ai, Atoms, Waves, Lightning, Pulse
Voice AI platform covering the full stack, pairing Atoms, a no code voice and chat agent builder handling conversations in more than twenty languages, with its own co optimized models, Lightning text to speech at roughly one hundred millisecond latency and Pulse speech to text with a sixty four millisecond first transcript.
Smallest AI was founded by Sudarshan Kamath and Akshat Mandloi and has raised $8 million in seed funding led by Sierra Ventures, months after a pre seed round led by 3one4 Capital. It builds many small, efficient models, each tuned for a narrow job and optimized to run together. The company says that in its first year it cut text to speech latency from roughly two seconds to one hundred milliseconds, dropped TTS cost from $0.20 to $0.01 per minute at scale, and moved from prototype to powering millions of enterprise calls each month.
The stack has three layers. Lightning is the text to speech line, with V3 hitting about one hundred millisecond latency across fifteen plus languages including Indic and European coverage, instant voice cloning from ten seconds of audio, and streaming over HTTP, SSE, or WebSocket. Pulse is streaming speech to text with a sixty four millisecond time to first transcript. Atoms is the agent platform on top.
It offers no code Single Prompt and Conversational Flow agents with an Agentic Graph Builder, knowledge bases, webhooks, API calls from within conversations, phone numbers, outbound campaigns and audiences, a web voice and chat widget, post call metrics, conversation logs, versioning, and testing. Developers also get Node.js and Python SDKs, a WebSocket SDK, mobile integrations, and an Atoms MCP server. The full agent pipeline stays under eight hundred milliseconds per turn, and the whole stack is available on AWS Marketplace under consolidated billing.
Smallest AI sells the model layer and the agent layer together for voice agents in support, sales and operations, with published usage pricing. Its multilingual TTS and STT models are also available as raw APIs.
Vendor details
Canonical URL
https://smallest.ai
Category
Voice agent
Subcategory
Full stack voice AI: Atoms agents plus Lightning TTS and Pulse STT
Funding status
Independent, founded by Sudarshan Kamath and Akshat Mandloi. Raised eight million dollars in seed funding led by Sierra Ventures, months after a pre seed led by 3one4 Capital, and reports powering millions of enterprise calls per month within a year of launch.
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
Atoms agents make API calls to outside systems mid call, run a pre call API request, and fire webhooks. Developers get REST APIs, SDKs, a WebSocket SDK, mobile SDKs and an Atoms MCP server that creates and configures agents and campaigns. Language models can be Smallest AI's own or any OpenAI compatible endpoint the customer brings, and the full stack is also sold through AWS Marketplace.
In practice
A support team ships a phone agent in an afternoon. A Single Prompt template opens with the prompt, voice and structure filled in, the team adjusts the voice, model and knowledge base, tests it over a web call and goes live on a real number.
A developer building a voice assistant needs speech that keeps up with conversation. Lightning V3.1 Pro returns speech in under one hundred milliseconds, and Pulse transcribes in real time with sixty four milliseconds of latency.
An operations team runs scheduled, high volume outbound campaigns from the same agent platform, with agents collecting information, completing transactions, and triggering downstream actions through API calls, while post call metrics and conversation logs feed review.
Sources & related URLs
Related / legacy domains
Research sources
Agentic Index coverage score
11.0 / 14 capabilities · 79%
| Integrations & Tool Calling | Full |
|---|---|
|
Agents carry API call tools that make HTTP requests to outside systems mid call, a pre call API request fetches context before the call connects, and the conversation log records the API calls and webhook triggers each call made. Tools are added, updated and removed per agent. SourceSmallest AI, docs.smallest.ai Atoms MCP available tools and Conversation logsread 2026-09-28 |
|
| Workflow Orchestration | Full |
|
Multi agent playbooks (SOPs with their own prompt, intent, auth level and tools) sit behind an intent router with a fallback playbook and mid call rerouting, Conversational Flow agents run node based flows, and the conversation log shows the routing path through multi agent setups. SourceSmallest AI, docs.smallest.ai Atoms MCP available tools and conversation logs pagesread 2026-09-28 |
|
| Knowledge Grounding & RAG | Full |
|
Knowledge bases hold uploaded PDFs and scraped web pages, are searched by semantic embeddings and injected into the model's context before the agent answers, and live apart from agents so content can be updated without touching agent configuration. SourceSmallest AI, docs.smallest.ai Knowledge Base overviewread 2026-09-28 |
|
| Human Oversight & Guardrails | Partial |
|
Agents hand calls to a person by cold or warm transfer when the prompt tells them to, with a private whisper to the destination on a warm transfer, and playbooks carry an auth level. There is no step where a person approves an agent action before it commits. SourceSmallest AI, docs.smallest.ai Call transfer and MCP available tools pagesread 2026-09-28 |
|
| Security, Identity & Governance | Full |
|
Organizations have Admin and Member roles, and only Admins manage the team, settings and billing. SSO is part of the Enterprise plan. Smallest AI states ISO 27001, SOC 2 Type 2, GDPR and HIPAA compliance and runs a security compliance portal. SourceSmallest AI, docs.smallest.ai Organization members, smallest.ai homepage and pricing pageread 2026-09-28 |
|
| Observability & Auditability | Full |
|
Each call's conversation log shows the timestamped transcript, the tool and function calls, API calls and webhook triggers, per turn latency, the routing path through multi agent setups, the recording, the cost breakdown and an event timeline, exportable as JSON or CSV. SourceSmallest AI, docs.smallest.ai Conversation logsread 2026-09-28 |
|
| Memory & State Persistence | Partial |
|
Session state lasts for the current call only, so anything that must survive between calls goes in the developer's own external store, such as Redis keyed by phone number. No platform memory across calls is provided. SourceSmallest AI, docs.smallest.ai State managementread 2026-09-28 |
|
| Deployment & Data Residency | Full |
|
On premises deployment is an option for the voice agent product, and the Enterprise plan offers dedicated infrastructure. No region list is published. SourceSmallest AI, smallest.ai homepage and pricing pageread 2026-09-28 |
|
| Prebuilt Agents, Templates & Packs | Partial |
|
Agents can start from a template gallery filtered by industry, direction and agent type, and a cookbook repository holds runnable agent templates and examples. The individual templates and the job each does are not named. SourceSmallest AI, docs.smallest.ai Quick start and Using cookbooksread 2026-09-28 |
|
| Triggers & Channel Coverage | Full |
|
Outbound campaigns start automatically at a scheduled date and time or through the API, with an audience and retry settings, and agents take calls on provisioned phone numbers. SourceSmallest AI, docs.smallest.ai Campaigns page and Atoms MCP available toolsread 2026-09-28 |
|
| Model Flexibility & Routing | Full |
|
Developers set each agent's language model, and bring your own model connects any endpoint that implements the OpenAI chat completions API, including self hosted Ollama, vLLM and LM Studio servers, beside Smallest AI's own Electron model. The customer chooses the model. SourceSmallest AI, docs.smallest.ai Bring your own modelread 2026-09-28 |
|
| APIs, SDKs & MCP Extensibility | Full |
|
An Atoms MCP server exposes tools that create, configure, duplicate and archive agents, manage their tools and playbooks, and create, start and pause campaigns, with drafts published to go live. Campaigns can also be created over the API, and REST, WebSocket and mobile SDKs are available beside it. SourceSmallest AI, docs.smallest.ai Atoms MCP available tools and Campaigns pageread 2026-09-28 |
|
| Testing, Debugging & Optimization | Partial |
|
Agents are tried by hand through web calls, telephony calls and chat before going live, agents can be duplicated to make variants, and post call analytics extract a summary and disposition with reasoning per call. There are no scored test cases, simulations or quality gates. SourceSmallest AI, docs.smallest.ai Evaluations and Call metrics pagesread 2026-09-28 |
|
| Browser & Computer Use | Not documented |
|
Agents act through API call tools, webhooks and telephony, with no browser, desktop or computer control. SourceSmallest AI, docs.smallest.ai Atoms pagesread 2026-09-28 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Product updates
A new POST /conversation/abort endpoint stops a voice agent call whether it is queued, ringing or live, for every conversation type. Queued calls are removed before dialing and live or ringing calls end within 5 seconds; aborted calls get a new aborted status, are billed only for answered time and are never redialed when a campaign resumes. It replaces the cancel endpoints, which only stopped queued calls and are now deprecated.
Bears on: Human approval / guardrails
View sourceSmallest AI added POST /waves/v1/auth/token so servers can issue temporary credentials for browser and mobile speech clients while retaining the API key. Tokens last 30 to 900 seconds, defaulting to 300, and authorize supported speech synthesis, transcription, speech to speech, and voice listing routes. They cannot authorize voice cloning, LLM chat completions, analytics, or further token creation; usage remains billed to the issuing key.
Bears on: Security / enterprise
View sourceSmallest AI shipped new MCP tools for latency summaries and prompt cache hit rates, reworked telephony so SIP trunks are managed as resources, and released SDK 5.5.0.
Bears on: MCP / tool calling / API
View sourcePricing
Pay as you go from $0.05 a minute, plus $0.01 a minute hosting; Enterprise tailored
Per agent minute, by model, plus a per minute hosting fee
What is public
Smallest AI charges for voice agents by the minute on pay as you go credits. Agent minutes cost $0.09 to $0.21 depending on the models selected, as low as $0.05, plus a flat $0.01 a minute for hosting. Pay as you go includes unlimited agents and 20 concurrent agents. Enterprise pricing is tailored.
Billing mechanics
There is no quota. Usage draws down prepaid credits at the per minute rate.
Cost watchouts
Telephony adds cost on agent minutes, and phone numbers rent at $10 a month each.
Variable cost rationale
Spend scales with call volume and call length, so long calls raise it faster than call counts.
Additional watchouts
Model choice moves the per minute rate by up to four times, so the bill follows the models actually run.
Sales call required
Mixed (some tiers require a call)
Free / trial
$10 in free credits for new accounts
Lowest paid plan
Pay as you go, from $0.05 a minute plus $0.01 a minute hosting
Cancellation / refund
Pay as you go needs no long term contract.
Support SLA / resale
Pay as you go has email and community support. Enterprise adds dedicated infrastructure, a 99.99% uptime SLA, forward deployed engineers, priority support and SSO.
Missing data
Smallest AI does not state Enterprise rates, the rate for each model combination, volume discount thresholds or a spend cap.
Related vendors
- AgentVoice — AgentVoice builds and runs inbound and outbound AI voice agents, and…
- aiola — Voice agents for field teams on Salesforce
- Aircall — Aircall's AI Voice Agents answer and place calls on Aircall's own…
- Airo by Ringover — Airo is Ringover's AI voice agent for inbound calls, outbound…
- Allo — Allo is a business phone system for small teams with an AI…
- Autocalls — Autocalls is a no code platform for AI voice agents that place and…
Alternatives to Smallest AI
Other products that do the work Smallest AI does, and what each one is.