> ## Documentation Index
> Fetch the complete documentation index at: https://docs.usenash.com/llms.txt
> Use this file to discover all available pages before exploring further.
> ## Agent Instructions
> Nash exposes two MCP servers. The docs MCP server at https://docs.usenash.com/mcp searches this documentation and needs no credentials. The Nash MCP server at https://mcp.usenash.com/mcp operates an organization's deliveries and needs a Nash API key. See https://docs.usenash.com/reference/build-with-ai.
> Use the Sandbox environment (https://api.sandbox.usenash.com/v1) for anything that creates or dispatches deliveries during development.
# Docs MCP server
> Connect Claude, Cursor, VS Code, or any MCP client to the Nash documentation so your AI assistant answers integration questions from the current docs.
The Nash documentation is available as a [Model Context Protocol](https://modelcontextprotocol.io) (MCP) server. Connect it to your AI assistant and the assistant can search this site and read any page while it helps you build — instead of guessing from training data or a stale web search.
| Field | Value |
| -------------- | ---------------------------------- |
| Server URL | `https://docs.usenash.com/mcp` |
| Transport | Streamable HTTP |
| Authentication | None — the documentation is public |
This server is read-only and reaches nothing but the documentation. To let an agent act on your deliveries — create orders, check live status, request refunds — connect the [Nash MCP server](/reference/mcp-integration) instead. The two are complementary, and a coding agent doing a Nash integration benefits from both.
## What the server exposes
The server gives an agent three tools, and the agent decides when to reach for each one based on the conversation:
* **Search** — find pages relevant to a query, returned as snippets with titles and links.
* **Read** — retrieve full pages as Markdown, navigate the site, and pull specific sections, including several pages in one call.
* **Submit feedback** — report a page that's wrong, outdated, or confusing. The report lands with the Nash docs team.
Everything in the site navigation is indexed, including the API reference, so an agent can look up a request body or a webhook payload the same way it looks up a concept page.
Agents and tools that support discovery can locate the server without being told where it is: `https://docs.usenash.com/.well-known/mcp` returns a JSON document with the server URL, and `https://docs.usenash.com/.well-known/mcp/server-card.json` adds the advertised tools and their schemas.
## Connect a client
The fastest route is the menu at the top of any docs page: **Connect to Cursor** and **Connect to VS Code** install the server in one click, and **Copy MCP server URL** puts the address on your clipboard for everything else.
1. Open [Connectors](https://claude.ai/settings/connectors) in the Claude settings.
2. Click **Add custom connector**.
3. Name it `Nash Docs` and set the URL to `https://docs.usenash.com/mcp`.
4. Click **Add**.
Click the attachments button (the plus icon), select **Nash Docs**, and ask your question.
```bash theme={"dark"}
claude mcp add --transport http nash-docs https://docs.usenash.com/mcp
```
Confirm the connection with `claude mcp list`. To share the server with everyone on a repository, commit it to a `.mcp.json` at the project root instead:
```json .mcp.json theme={"dark"}
{
"mcpServers": {
"nash-docs": {
"type": "http",
"url": "https://docs.usenash.com/mcp"
}
}
}
```
Click **Connect to Cursor** in the menu at the top of this page, or add the server by hand:
1. Open the command palette (Cmd + Shift + P, or Ctrl + Shift + P on Windows) and search for **Open MCP settings**.
2. Click **Add custom MCP** to open `mcp.json`.
3. Add the server:
```json mcp.json theme={"dark"}
{
"mcpServers": {
"nash-docs": {
"url": "https://docs.usenash.com/mcp"
}
}
}
```
Click **Connect to VS Code** in the menu at the top of this page, or create `.vscode/mcp.json` in your workspace:
```json .vscode/mcp.json theme={"dark"}
{
"servers": {
"nash-docs": {
"type": "http",
"url": "https://docs.usenash.com/mcp"
}
}
}
```
Add the server to `~/.codex/config.toml`:
```toml ~/.codex/config.toml theme={"dark"}
[mcp_servers.nash-docs]
url = "https://docs.usenash.com/mcp"
```
Any client that speaks MCP over streamable HTTP can connect. Give it the URL `https://docs.usenash.com/mcp` with no credentials. The search and read tools are read-only and idempotent, and they say so in their MCP annotations, so clients that gate tool calls on those hints can run them without a prompt.
## Try it
Once connected, ask something the assistant would otherwise have to guess at:
```text theme={"dark"}
Using the Nash docs, show me how to create an order with a scheduled dropoff window
and then select the cheapest quote. Include the exact request bodies.
```
A well-behaved agent searches the docs, reads the [Create Order](/api-reference/order/create-order) and [Select Quote](/api-reference/order/select-quote) pages, and answers from what's there. If it reports a page as wrong along the way, that feedback reaches us.
## Limits and scope
* The server searches the current published documentation. Anything not on this site — your organization's configuration, your orders, Portal-only settings — is out of its reach. Use the [Nash MCP server](/reference/mcp-integration) for live data.
* Requests are rate limited per client and per site. A client that hits the limit gets an HTTP `429` and should back off; limits reset on a rolling hourly window.
* Connecting several documentation servers to one assistant makes every search more expensive. Keep only the ones relevant to what you're building connected.
## Related
Let an agent operate deliveries — create, track, refund — with your API key.
Paste-able references for assistants that can't connect to MCP.