Skip to content
J·BJon BishopIndependent growth marketer

Artificial intelligence · llms.txt file breakdown

Sarvam AI Developer Documentation llms.txt

Build with Sarvam AI. Speech-to-text, text-to-speech, translation, chat, and document intelligence APIs for 22 Indian languages, plus no-code platforms for voice agents, coding agents, Cowork, and Content Studio.

The actual file

Source Markdown

The saved llms.txt text, displayed as source. The file belongs to its publisher; the live endpoint may change.

Open saved text ↗

Complete saved Markdown shown below.

# Sarvam AI Developer Documentation

> Build with Sarvam AI. Voice Agents: a managed platform for production voice agents in 11 Indian languages, from ₹3 per minute. Plus speech-to-text, text-to-speech, translation, chat, and document intelligence APIs for 22 Indian languages, and no-code platforms for coding agents, Cowork, and Content Studio.

Indus: Full Stack AI Platform

# AI for all from India.

Speech, text, and document models trained for 22 Indian languages and English, plus
no-code platforms for voice agents, coding agents, Cowork, and Content Studio. Create a key
and make your first call in a minute.

[Get started](/api/getting-started/quickstart)

[Browse the API reference](/api-reference/introduction)

Native support for हिन्दी বাংলা தமிழ் తెలుగు मराठी ગુજરાતી ಕನ್ನಡ മലയാളം ਪੰਜਾਬੀ ଓଡ଼ିଆ অসমীয়া اردو नेपाली कोंकणी کٲشُر سنڌي संस्कृतम् संताली মৈতৈলোন্ बर' मैथिली डोगरी English

### Call the models

REST, WebSocket, and batch endpoints for transcription, speech synthesis, translation,
chat, and document extraction. Official Python and JavaScript SDKs.

[Quickstart](/api/getting-started/quickstart)

[SDKs](/api/getting-started/sdks)

### Build agents

Design, deploy, and monitor conversational agents across phone and chat in 11 Indian
languages, hand recurring work to agents wired into Gmail, Slack, and Notion, or deploy coding agents for development tasks.

[Voice Agents](/voice-agents/overview)

[Cowork](/cowork/overview)

[Coding Agents](/api/sarvam-code)

### Work without code

Dub video, clone voices, generate speech, and turn scans and handwritten forms into
structured data, all from the dashboard with no integration required.

[Content Studio](/creative/studio-overview)

[Doc Agents](/docai/getting-started/overview)

## Explore the platform

Seven entry points, one account. Start wherever your problem lives.

[Model APIs](/api/getting-started/welcome)

Speech, text, chat, document intelligence, and document-to-document translation behind a single API key. Build it into your own product.

[Voice Agents](/voice-agents/overview)

Build, deploy, and monitor voice and chat agents. Bring your own telephony or rent a number from Sarvam.

[Doc Agents](/docai/getting-started/overview)

Extract fields and digitise documents (invoices, KYC, land deeds, handwritten manuscripts) into JSON, CSV, Markdown, or DOCX.

[Document Translation](/api/api-guides-tutorials/doc-translation/overview)

Translate whole PDF and Office files into up to 12 Indian languages per job — layout preserved, document-to-document (PDF stays PDF, DOCX stays DOCX).

[Content Studio](/creative/studio-overview)

Dub video, clone voices, generate speech, and transcribe audio from the browser.

[Cowork](/cowork/overview)

An agentic workspace for real work: email triage, morning briefings, internal helpdesk, lead follow-up.

[Coding Agent](/api/sarvam-code)

An agentic coding assistant for your terminal and IDE, built on Sarvam models for development tasks.

## Build with the APIs

Guides and tutorials for each capability, from a first request to production
patterns.

[All API guides →](/api/getting-started/welcome)

[Speech-to-Text Saaras v3](/api/api-guides-tutorials/speech-to-text/overview)

Transcribe, translate, and caption audio in real time or in batch, across code-mixed and regional speech.

[Text-to-Speech Bulbul v3](/api/api-guides-tutorials/text-to-speech/overview)

Generate natural, expressive speech in Indian languages with control over pitch, pace, and speaker.

[Chat Completion Sarvam-105B](/api/api-guides-tutorials/chat-completion/overview)

Build conversational AI on a flagship model tuned for Indic reasoning, with a 128K context window.

[Text translation Mayura](/api/api-guides-tutorials/text-processing/overview)

Translate, transliterate, and detect languages in short text — not whole files. Use Document Translation for PDF and Office exports.

[Document Translation](/api/api-guides-tutorials/doc-translation/overview)

Translate PDF, Word, Excel, PowerPoint, and HTML into up to 12 languages per job while keeping layout and formatting intact.

[Document AI Sarvam Vision](/api/api-guides-tutorials/document-intelligence/overview)

Turn PDFs, scans, and handwriting into structured, searchable data: whole documents or just the fields you define.

[Dubbing Beta](/api/api-guides-tutorials/dubbing/overview)

Localize video and audio into Indian languages with speaker voice cloning, tone control, and multi-format exports.

[Instant Voice Cloning Beta](/api/api-guides-tutorials/voice-cloning/overview)

Generate speech in a cloned voice from a short reference clip, in 13 Indian languages, with cross-lingual cloning and built-in quality control.

## Models

Every model is trained and evaluated on Indian languages, accents, and code-mixed
usage, not adapted from an English-first baseline.

[Compare models →](/api/getting-started/models)

[Saaras v3 Speech-to-text](/api/getting-started/models/saaras)

23 languages (22 Indian + English), with transcribe, translate, verbatim, translit, and codemix output modes.

[Bulbul v3 Text-to-speech](/api/getting-started/models/bulbul)

Natural voices for 11 languages (10 Indian + English), with customizable pitch, pace, and speaker.

[Sarvam-105B Chat](/api/getting-started/models/sarvam-105b)

Flagship 105B-parameter chat model: the highest quality Indian language understanding, reasoning, and generation.

[Mayura Translation](/api/getting-started/models/mayura)

High-quality translation between 11 languages (10 Indian + English) with context preservation.

[Sarvam Vision Document intelligence](/api/getting-started/models/sarvam-vision)

Document model with accurate OCR and structured output across 23 languages.

## Integrations

Drop Sarvam into the frameworks and telephony you already run.

[All integrations →](/api/integration/build-voice-agent-with-live-kit)

[LiveKit](/api/integration/build-voice-agent-with-live-kit)

Real-time voice agents with Saaras and Bulbul in the pipeline.

[Pipecat](/api/integration/build-voice-agent-with-pipecat)

Build a real-time voice agent wired end to end to Sarvam models.

[Vapi](/api/integration/build-voice-agent-with-vapi)

Run a Vapi agent on Saaras v3, Sarvam-105B, and Bulbul v3.

[Twilio](/api/integration/build-voice-agent-with-twilio)

Phone voice agents and WhatsApp bots over Twilio.

[Exotel](/api/integration/build-voice-agent-with-exotel)

Phone voice agents on Exotel telephony, via Pipecat.

[LangChain](/api/integration/langchain)

The official langchain-sarvam package: streaming, tools, structured output.

[Vercel AI SDK](/api/integration/vercel-ai-sdk)

Chat, speech, and translation through the sarvam-ai-sdk provider.

[n8n](/api/integration/n8n)

Automate Indian-language speech and chat with the community node.

## Cookbook

End-to-end projects you can run today, with the full source in each guide.

[All recipes →](/api/cookbook)

[Document translation API](/api/api-guides-tutorials/doc-translation/guides/translate-a-pdf)

Upload a PDF or Office file, poll job status, and download layout-preserving translations for multiple Indian languages.

[Call analytics pipeline](/api/cookbook/guides/call-analytics-pipeline)

Transcribe support calls, then summarise, tag, and score them into a reporting pipeline.

[WhatsApp bot](/api/cookbook/guides/whats-app-bot)

Wire Sarvam speech and chat into a WhatsApp bot that answers in the user's own language.

[Live video transcription](/api/cookbook/guides/live-video-transcription)

Real-time transcription and translation with the streaming STT API, Flask-SocketIO, and the Web Audio API.

[Podcast generator](/api/cookbook/guides/podcast-generator)

Turn a script into a multi-speaker podcast episode with distinct Bulbul voices.

## What's new

[Full changelog →](/changelog)

[October 2026 Bulbul V3 End of Life for Voice Agents](/voice-agents/build/voice-language)

Bulbul V3 reaches end of life on Voice Agents on November 10, 2026. Switch your agents to Bulbul V4 before then.

Read the changelog →

[October 2026 Updated Pricing for Language Models](/api/getting-started/pricing)

New per-token prices for Sarvam 105B, Gemma 4, GLM-5.3, and DeepSeek V4.1 Flash, including lower input and output prices for Sarvam 105B and Gemma 4.

[October 2026 Instant Voice Cloning in the sarvamai SDKs](/api/api-guides-tutorials/voice-cloning/overview)

Python and TypeScript SDKs expose Instant Voice Cloning (beta) with the same convert / convertStream / WebSocket shape as Bulbul TTS.

[October 2026 Session Affinity for Chat Completion V2](/api/api-guides-tutorials/chat-completion/overview#session-affinity-for-multi-turn-conversations)

Send an optional `x-session-affinity` header so multi-turn `gemma4` conversations reuse cached prompt context for lower latency.

## Resources

Get started

* [Quickstart](/api/getting-started/quickstart)
* [Authentication](/api-reference/authentication)
* [SDKs & libraries](/api/getting-started/sdks)
* [Building for Indian languages](/api/getting-started/building-for-india)

Operate

* [Pricing](/api/getting-started/pricing)
* [Credits & rate limits](/api/getting-started/ratelimits)
* [Errors & troubleshooting](/api/getting-started/errors-troubleshooting)
* [API status](https://status.sarvam.ai/)

Tools

* [MCP server](/api/developer-tools/mcp)
* [Agent Skills](/api/developer-tools/agent-skills)
* [Meta prompt guide](/api-reference/metaprompt)
* [Self-hosted deployments](/api/self-hosted/introduction)

Community

* [Dashboard](https://dashboard.sarvam.ai/)
* [Discord](https://discord.com/invite/5rAsykttcs)
* [GitHub](https://github.com/sarvamai/sarvam-ai-cookbook)
* [Talk to us](/api/getting-started/help)

> Build with Sarvam AI. Voice Agents: a managed platform for production voice agents in 11 Indian languages, from ₹3 per minute. Plus speech-to-text, text-to-speech, translation, chat, and document intelligence APIs for 22 Indian languages, and no-code platforms for coding agents, Cowork, and Content Studio.

## Instructions for AI Agents

- For clean Markdown of any page, append `.md` to the page URL
- For section-specific indexes, append `/llms.txt` to any section URL
- For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.sarvam.ai/_mcp/server

## Products

- [Model APIs](https://docs.sarvam.ai/api/llms.txt): Speech, text & chat model APIs
  - [Welcome](https://docs.sarvam.ai/api/getting-started/welcome.md)
  - [Quickstart](https://docs.sarvam.ai/api/getting-started/quickstart.md)
  - [SDKs & Libraries](https://docs.sarvam.ai/api/getting-started/sdks.md)
  - [Building for Indian Languages](https://docs.sarvam.ai/api/getting-started/building-for-india.md)
  - [Saaras](https://docs.sarvam.ai/api/getting-started/models/saaras.md)
- [Voice Agents](https://docs.sarvam.ai/voice-agents/llms.txt): Build, deploy & monitor voice agents
  - [Introduction](https://docs.sarvam.ai/voice-agents/introduction.md)
  - [Why Voice Agents](https://docs.sarvam.ai/voice-agents/why-voice-agents.md)
  - [Overview](https://docs.sarvam.ai/voice-agents/overview.md)
  - [Quickstart](https://docs.sarvam.ai/voice-agents/quickstart.md)
  - [MCP Server](https://docs.sarvam.ai/voice-agents/mcp.md)
- [Doc Agents](https://docs.sarvam.ai/docai/llms.txt): Extract & digitise documents
  - [Overview](https://docs.sarvam.ai/docai/getting-started/overview.md)
  - [Extract structured fields](https://docs.sarvam.ai/docai/how-to/extract-fields-from-a-document/extract-structured-fields.md)
  - [Extract as CSV/Excel](https://docs.sarvam.ai/docai/extract-as-csv-excel.md)
  - [Create a reusable config](https://docs.sarvam.ai/docai/how-to/extract-fields-from-a-document/create-a-reusable-config.md)
  - [Digitise a document](https://docs.sarvam.ai/docai/how-to/digitise-a-document.md)
- [Content Studio](https://docs.sarvam.ai/creative/llms.txt): Generate voice & media content
  - [Overview](https://docs.sarvam.ai/creative/studio-overview.md)
  - [Creating a Dub](https://docs.sarvam.ai/creative-dubbing-create.md)
  - [Editing a Dub](https://docs.sarvam.ai/creative-dubbing-edit.md)
  - [Dubbing Best Practices](https://docs.sarvam.ai/creative-dubbing-best-practices.md)
  - [Generating Speech](https://docs.sarvam.ai/creative-tts-generate.md)
- [Cowork](https://docs.sarvam.ai/cowork/llms.txt): Build & deploy AI agents for real work
  - [Overview](https://docs.sarvam.ai/cowork/overview.md)
  - [Quickstart](https://docs.sarvam.ai/cowork/quickstart.md)
  - [Chat](https://docs.sarvam.ai/cowork/build/chat.md)
  - [Create custom agents](https://docs.sarvam.ai/cowork/build/creating-an-agent.md)
  - [Connectors](https://docs.sarvam.ai/cowork/build/connectors.md)
- [API Reference](https://docs.sarvam.ai/api-reference/llms.txt): Explore REST & WebSocket endpoints
  - [Introduction](https://docs.sarvam.ai/api-reference/introduction.md)
  - [Authentication](https://docs.sarvam.ai/api-reference/authentication.md)
  - [Beta APIs and Models](https://docs.sarvam.ai/api-reference/beta-apis.md)
  - [Meta Prompt Guide](https://docs.sarvam.ai/api-reference/metaprompt.md)
  - [Introduction](https://docs.sarvam.ai/api-reference/conversations.md)
- [Changelog](https://docs.sarvam.ai/changelog/llms.txt): Track API and product updates
  - [October 22, 2026](https://docs.sarvam.ai/changelog/2026/10/22.md)
  - [October 16, 2026](https://docs.sarvam.ai/changelog/2026/10/16.md)
  - [October 14, 2026](https://docs.sarvam.ai/changelog/2026/10/14.md)
  - [October 12, 2026](https://docs.sarvam.ai/changelog/2026/10/12.md)
  - [October 10, 2026](https://docs.sarvam.ai/changelog/2026/10/10.md)

Measured from the saved file

Linked document

UTF-8 text size
13.5 KB13871 bytes
Token count
3524o200k_base encoding
Markdown links
10996 distinct URLs
Fenced code blocks
0313 lines in the file

Size measures the captured UTF-8 text, not compressed transfer size. Tokens are counted with a named tokenizer; other models can count differently. Neither measure establishes context compatibility or answer quality.

Project H1
Sarvam AI Developer Documentation

Opening blockquote

Build with Sarvam AI. Voice Agents: a managed platform for production voice agents in 11 Indian languages, from ₹3 per minute. Plus speech-to-text, text-to-speech, translation, chat, and document intelligence APIs for 22 Indian languages, and no-code platforms for coding agents, Cowork, and Content Studio.

Context architecture

How the content is divided

9 H2 sections organize 109 parsed Markdown links. 35 links point to .md or .markdown paths; 4 link occurrences use another hostname. Linked destinations have not been recursively validated.

H2 sectionLinksUTF-8 bytesTokens
Explore the platformLine 4871198277
Build with the APIsLine 8091711389
ModelsLine 11961014240
IntegrationsLine 14691055271
CookbookLine 1846957227
What's newLine 21051025270
ResourcesLine 232161193323
Instructions for AI AgentsLine 264025666
ProductsLine 2704233101010

Nested indexes

7 distinct links point to other llms.txt paths. These are routes to more indexes, not proof their contents were included in this file.

Link counts exclude parsed fenced and inline code and resolve relative destinations against the final endpoint. Section sizes cover the text below each H2 until the next H2.

Embedded material

Code, installation and API examples

Fenced blocks
0 closed
Declared languages
No fenced blocks detected
Package-install commands
0 detected in fenced or inline code
Authentication examples
0 fenced lines match API-key, Authorization, Bearer, OAuth or access-token patterns
HTTP request examples
0 fenced lines match request-method patterns
Authentication links
0 links have authentication-related labels

Embedded example text and links to further documentation are counted separately. A code block does not establish that a sample runs or that an authentication flow succeeds.

Publisher prioritization

The Optional section

No H2 section named “Optional” was detected in this file.

The community proposal uses Optional by convention for secondary links that can be skipped for a shorter context. This label does not establish that everything outside it is critical, or that the material inside it is deprecated.

Machine-readable routes

MCP references and configuration text

MCP-related text was detected in the captured file. 2 links have MCP-related labels or paths. 0 fenced lines match mcpServers, mcp_servers or an mcp add command.

These are text and link observations. No MCP server connection, authentication or tool call was tested.

Index and bundle

llms-full.txt companion

Verified text companion

Text companion verified

https://docs.sarvam.ai/llms-full.txt ↗
Discovery
Conventional sibling path checked
HTTP response
206 · text/plain; charset=utf-8
Text size
13.5 KB · 13875 bytes (server-reported)
Token count
3510 o200k_base tokens
Retrieval scope
Complete body retrieved

A verified text response establishes a companion at the checked URL. It does not prove the bundle contains all product documentation, that generation is automated, or that it fits an assistant’s context window.

Change evidence

What can be said about freshness

This capture establishes a baseline. Update frequency cannot be determined from one observation.

Linked release and migration material

7 links have release, changelog, deprecation or migration terms in their labels or paths. These provide places to investigate product changes separately.

File changes can reveal documentation edits or navigation changes. They do not by themselves measure product release velocity, accuracy, agent adoption or AI visibility.

Capture identity and measurement method
Requested endpoint
https://docs.sarvam.ai/llms.txt
Final endpoint
https://docs.sarvam.ai/llms.txt
HTTP status
200
Content type
text/plain; charset=utf-8
Text retrieval
Firecrawl literal text capture
Tokenizer
tiktoken 0.14.0 · o200k_base
SHA-256 of saved UTF-8 text
b251a61b1d3e548b7d1716a8abadee9c5c316eb3682ddd489dd21b0d0c6b822f

Counts use a deterministic parser, not a spec-compliance grade. Unusual Markdown and indented code can fall outside its recognized patterns. See the llms.txt proposal and tokenizer implementation.

Related tools and resources

Prefer to write? Send Jon a message.