> 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.

# Create custom agents

**Agents** are for recurring work. Each agent has a name, a set of instructions (system prompt) describing its job, the **Connectors** it can use, the **Skills** it knows, and the **Knowledge Bases** it draws on. Build it once and it produces consistent results every time, on demand or on a schedule.

<img src="https://fdr-prod-docs-files-public.s3.us-east-1.amazonaws.com/sarvam-api-docs.docs.buildwithfern.com/b77b5ef4cfc0c50b8355000795f2ba7a9f4bc8bf6b8b9a00e73d214ec540bfa9/sarvam-work-agents/images/agents.png?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Content-Sha256=UNSIGNED-PAYLOAD&X-Amz-Credential=AKIA6KXJSKKNFOCF7G4B%2F20260730%2Fus-east-1%2Fs3%2Faws4_request&X-Amz-Date=20260730T094647Z&X-Amz-Expires=604800&X-Amz-Signature=64ae3bba4224028e46eab839f99ae363c3a89b0e010fcc985ace53a974b22677&X-Amz-SignedHeaders=host&x-amz-checksum-mode=ENABLED&x-id=GetObject" alt="Sarvam Work Agents Agents page with a 'Build an agent for your work' prompt box and a grid of saved agents including Docs, PPT maker, Slack Tool, and Equity Researcher." />

## Use the Agent Builder

To create one, click **Agents** from the sidebar on the Home page. This opens the **Agent Builder**, where you describe what the agent should do in plain English. The Agent Builder drafts the instructions (system prompt), suggests the connectors and external systems the agent will need, and helps you configure it. You can also upload files as context: documents, briefs, or reference material that the Agent Builder incorporates into the setup.

From the Agent Builder you can 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.

## Test in the playground

Inside the **Agent Builder**, switch the **Builder / Playground** toggle to **Playground** and chat with the agent to test how it responds. Toggle back to **Builder** to make adjustments, then save when you are happy with the results. Refine its instructions, connectors, skills, and knowledge until the output is what you expect.

## Sub-agents

For complex jobs, an agent can spin up **sub-agents**: smaller, focused workers that tackle individual pieces of the task in parallel for faster results.

## Share, schedule, and deploy

Agents improve as your team uses them. When ready, share them, schedule them, or deploy them to Slack.

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

Run the agent automatically on a repeating schedule or once at a set time.

#### [Connectors](/work-agents/build/connectors)

Give the agent access to Slack and the other tools it needs to act.