> ## Documentation Index
> Fetch the complete documentation index at: https://docs.orisu.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Quickstart

> Add the Orisu MCP connector to Claude or ChatGPT and run your first agent from chat.

The connector takes two minutes to add. Once it's in, your chat client can drive Orisu end-to-end.

<Snippet file="snippets/mcp-endpoints.mdx" />

## In Claude

<Steps>
  <Step title="Open Connectors">
    Claude → Settings → Connectors → **Add custom connector**.
  </Step>

  <Step title="Paste the Orisu MCP endpoint">
    Paste this URL into the connector setup field:

    ```
    https://api.orisu.ai/mcp
    ```

    <Tip>
      Your Orisu dashboard's **Settings → Connectors → MCP** page always shows the live endpoint for your account. Copy from there if this URL ever lags a deploy or you're on a non-production environment.
    </Tip>

    Click Continue.
  </Step>

  <Step title="Sign in to Orisu">
    A Orisu login window opens (OAuth). Sign in with the account you normally use. The first time, you'll see a consent screen listing the scopes the connector needs. Approve to continue.
  </Step>

  <Step title="Pick a workspace">
    If your Orisu account belongs to multiple workspaces, you'll be asked which one to bind the connector to. All subsequent tool calls run against that workspace.
  </Step>

  <Step title="Try it">
    Open a new chat and ask: *"List my recent Orisu runs."* Claude calls `list_runs`, the connector returns your latest runs, and Claude summarises them. That's the loop working.
  </Step>
</Steps>

<Tip>
  The first message in a fresh chat triggers the model to load the [orisu-workflows skill](/mcp/tools#workflow-skill), the canonical playbook that teaches it the patterns. After that, follow-up turns are fast because the playbook is already in context.
</Tip>

## In ChatGPT

ChatGPT's connector setup follows the same shape. Paste the same MCP endpoint URL, complete the OAuth flow, and you're in. Refer to ChatGPT's connector documentation for the latest UI flow.

## What to ask first

Good first prompts that exercise the full surface:

* **Discovery:** *"What Orisu agents do I have?"* → `search` + `orisu://agents`
* **Building:** *"Create a workflow that takes a photo and produces a 3-second video clip with motion."* → `create_agent` + `validate_graph` + canvas render
* **Running:** *"Trigger my last agent with this prompt and wait for it to finish."* → `trigger_run` + `wait_for_run` + asset URLs
* **Iterating:** *"That run failed — what happened?"* → reads `orisu://runs/{id}` and explains the error

## Next steps

<Columns cols={2}>
  <Card title="Tool reference" icon="wrench" href="/mcp/tools">
    What each of the 20+ tools does.
  </Card>

  <Card title="In-chat canvas" icon="layout-dashboard" href="/mcp/canvas">
    Understand the live canvas, when it appears, what's interactive.
  </Card>

  <Card title="API keys" icon="key" href="/mcp/authentication#api-keys">
    For scripts and server-to-server, no OAuth required.
  </Card>

  <Card title="Error codes" icon="circle-alert" href="/mcp/errors">
    What each error means and how to recover.
  </Card>
</Columns>