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

# Use Avatcado with AI tools

> Search these docs from an AI assistant over MCP, and the machine-readable files for assistants and crawlers

# Use Avatcado with AI tools

These docs can be read by AI assistants and editors in three ways: through an MCP server that searches the documentation, through plain-text index files, and through the OpenAPI spec.

> [!WARNING]
>
> The MCP server searches documentation only. It does not validate VAT numbers and it needs no API key. To validate a number, call [`GET /v1/validate`](https://docs.avatcado.com/api-reference/vat-validation/validate-a-vat-number) or use an [SDK](#sdks).

## Documentation MCP server

Endpoint:

```text theme={null}
https://docs.avatcado.com/mcp
```

It is a remote server over HTTP, with no local install. An assistant that connects to it can look up how Avatcado works and read the pages it finds, so answers about error codes, parameters or tiers come from the current docs.

| Tool | What it does |
| - | - |
| `search_avatcado` | Searches the documentation and returns matching pages with titles and links. Takes a `query` string. |
| `query_docs_filesystem_avatcado` | Runs read-only shell-style commands (`rg`, `grep`, `ls`, `tree`, `cat`, `head`, `find`, `jq`) against a virtual copy of the docs pages and OpenAPI specs. A page at `/caching` is read as `/caching.mdx`. |

Both tools are read-only. Nothing runs on your machine, and the filesystem holds only the Avatcado docs.

### Add it to an assistant or editor

Any client that supports remote MCP servers can use it. Look for the setting that adds a remote or HTTP MCP server, and enter the endpoint URL above as the server URL. The exact menu and config file differ per client, so follow your client's own MCP documentation for that step.

Once connected, ask a question that the docs answer:

* "How do I send a consultation number with a validation request?"
* "What does the `upstream_member_state_unavailable` error mean?"
* "Which plans include batch validation?"

### Check the endpoint from a terminal

This request lists the two tools. It does not need an API key:

```bash theme={null}
curl -s -X POST https://docs.avatcado.com/mcp \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'
```

The response lists `search_avatcado` and `query_docs_filesystem_avatcado` with their input schemas.

## Files for assistants and crawlers

| File | URL | Contents |
| - | - | - |
| Docs `llms.txt` | [docs.avatcado.com/llms.txt](https://docs.avatcado.com/llms.txt) | Every documentation, error and API reference page with a one-line description |
| Site `llms.txt` | [www.avatcado.com/llms.txt](https://www.avatcado.com/llms.txt) | Summary of the API, coverage, pricing and the main pages of the marketing site |
| Site `llms-full.txt` | [www.avatcado.com/llms-full.txt](https://www.avatcado.com/llms-full.txt) | The full text of every developer guide and every VIES error reference page |
| OpenAPI spec | [docs.avatcado.com/openapi/public.yaml](https://docs.avatcado.com/openapi/public.yaml) | Every public endpoint, parameter and response schema |

Give an assistant the OpenAPI spec URL when you want it to generate a client or write requests that match the schema. Give it the `llms.txt` URL when you want it to find the right page first.

## SDKs

The official SDKs wrap the API and return typed results:

* Node.js: `npm install @avatcado/node` ([npm](https://www.npmjs.com/package/@avatcado/node))
* Python: `pip install avatcado` ([PyPI](https://pypi.org/project/avatcado/))

See the [Quickstart](https://docs.avatcado.com/quickstart.md) for a first request in curl, Node.js and Python.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.