> For clean Markdown of any page, append `.md` to the page URL.
> For a complete documentation index, see https://docs.sarvam.ai/llms.txt.
> For full documentation content in one file, see https://docs.sarvam.ai/llms-full.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.sarvam.ai/_mcp/server.

# Quickstart

Run your first task, then turn it into a reusable agent, connect a tool, and schedule it.

## Start working

Open [platform.sarvam.ai/indus](https://platform.sarvam.ai/indus) and sign in. You land on **Home**, with a prompt bar in the centre and a sidebar for **Connectors**, **Skills**, **Agents**, **Scheduled Tasks**, and **Knowledge Bases**.

#### [Open Sarvam Work Agents](https://platform.sarvam.ai/indus)

Type a task in the prompt bar to start immediately, or open Agents to build something reusable.

## Step 1: Run a task in Chat

1. On **Home**, type a task into the prompt bar (for example, "Summarise this report and pull out the key numbers").
2. Attach any files or images for context: a PDF to summarise, a spreadsheet to analyse, a document to draft from.
3. Send it and watch the response come back. The session keeps context, so you can follow up without re-explaining.

See [Chat](/work-agents/build/chat) for more on ad-hoc tasks.

## Step 2: Turn it into an agent

If it's something you'll do again, build an agent for it.

1. Click **Agents** from the sidebar on the Home page to open the **Agent Builder**.
2. Describe what the agent should do in plain English. The Agent Builder drafts the instructions, suggests the connectors it will need, and helps you configure it.
3. Optionally upload files as context for the setup.
4. In the Agent Builder, add or remove the **Connectors**, **Skills**, and **Knowledge Bases** the agent can use, so it has exactly the tools, know-how, and context it needs.

See [Create custom agents](/work-agents/build/creating-an-agent).

## Step 3: Test in the playground

1. In the **Agent Builder**, switch the **Builder / Playground** toggle to **Playground**.
2. Chat with the agent to test how it responds, and review the result.
3. Toggle back to **Builder** to adjust the instructions, connectors, skills, or knowledge, then save when you're happy.

See [Create custom agents](/work-agents/build/creating-an-agent) for the full build-and-test loop.

## Step 4: Connect a tool

To let your agent act in the systems you use, add a Connector.

1. Open **Connectors** and pick a service (start with Gmail or Calendar).
2. Complete the secure OAuth flow to authorise it. This adds the connector to your **Chat** tab.

See [Connectors](/work-agents/build/connectors).

## Step 5: Run it on a schedule

1. Open **Scheduled Tasks** and point a schedule at your agent.
2. Choose when it should run (for example, every morning).
3. The work is done before you need it, no manual trigger required.
4. Or just tell your agent to schedule the task in plain language, and it sets up the schedule for you.

See [Scheduled Tasks](/work-agents/build/scheduled-tasks).