# Build with Agents

> For the complete documentation index, see [llms.txt](https://docs.banxa.com/llms.txt). Append `.md` to any page URL for its markdown version.


This documentation is built to be read by agents as well as by people. Everything
below is read-only, and needs no credentials. It provides the information for your
agent to help you build your integration faster.

Why it matters
An agent working from a web search may hallucinate and invent endpoints.
Pointing it at these documentations means it reads the actual parameter
names, the actual error codes, and the actual auth scheme.

## MCP server

These docs run a [Model Context Protocol](https://modelcontextprotocol.io/)
server. An agent connected to it searches and reads the documentation directly
rather than scraping rendered HTML.

```bash
https://docs.banxa.com/mcp
```

It exposes six read-only tools:

| Tool | What it returns |
|  --- | --- |
| `search` | Passages matching a query, across all three products |
| `list-apis` | The documented APIs and what each is for |
| `get-endpoints` | Every endpoint of one API, with method and path |
| `get-endpoint-info` | One endpoint in full (parameters, schemas, security, examples) |
| `get-security-schemes` | How to authenticate against one API |
| `get-full-api-description` | The complete OpenAPI description |


The transport is streamable HTTP. Clients must send
`Accept: application/json, text/event-stream`; a request advertising only
`application/json` is rejected with `406 Not Acceptable`.

### Connect your client

```bash Claude Code
claude mcp add --transport http banxa-docs https://docs.banxa.com/mcp
```

```json Cursor
// .cursor/mcp.json
{
  "mcpServers": {
    "banxa-docs": {
      "url": "https://docs.banxa.com/mcp"
    }
  }
}
```

```json VS Code
// .vscode/mcp.json
{
  "servers": {
    "banxa-docs": {
      "type": "http",
      "url": "https://docs.banxa.com/mcp"
    }
  }
}
```

Once connected, ask the agent questions:

* which Banxa product should I use if I already run my own KYC,
* what does `POST /eapi/v0/ramps` accept,
* what does error `40003` mean


A machine-readable description of the server, including the tool list, is available
at [`https://docs.banxa.com/.well-known/mcp/server-card.json`](https://docs.banxa.com/.well-known/mcp/server-card.json).

## Skills

Two skills are published for coding agents. Each is a self-contained instruction
document an agent can fetch and follow, rather than a page it has to interpret.

| Skill | Use it when |
|  --- | --- |
| [`choose-banxa-integration`](https://docs.banxa.com/.well-known/agent-skills/choose-banxa-integration/SKILL.md) | Deciding between Native API, Hosted Checkout, or other future integration patterns |
| [`banxa-api-authentication`](https://docs.banxa.com/.well-known/agent-skills/banxa-api-authentication/SKILL.md) | Signing requests, and debugging `401`s or nonce collisions |


The index is at
[`/.well-known/agent-skills/index.json`](https://docs.banxa.com/.well-known/agent-skills/index.json),
with a SHA-256 digest per skill so a client can verify it fetched the document
the index describes.

## Every page as markdown

Append `.md` to any documentation URL to get the markdown source instead of
rendered HTML.

```bash
curl https://docs.banxa.com/products/native-api/docs/getting-started/authentication.md
```

## llms.txt

Banxa provides [llms.txt](https://llmstxt.org/) for LLMs to consume our documentation on their own.

A condensed version of the documentation is available at [`/llms.txt`](https://docs.banxa.com/llms.txt).

The full corpus is available at [`/llms-full.txt`](https://docs.banxa.com/llms-full.txt).

## API catalog

[`https://docs.banxa.com/.well-known/ai-catalog.json)`](https://docs.banxa.com/.well-known/api-catalog) is an
[RFC 9727](https://www.rfc-editor.org/rfc/rfc9727) catalog listing every Banxa
API with its reference documentation and status endpoint, so a client can
enumerate what exists before fetching anything.

## In-browser tools

WebMCP is an emerging standard
While it is  still stabilising in Chrome, please treat this feature as experimental

Pages on this site register [WebMCP](https://webmachinelearning.github.io/webmcp/)
tools, so a browser-resident agent can search these docs and look up endpoints
without leaving the page.

## Crawling policy

[`/robots.txt`](https://docs.banxa.com/robots.txt) names the AI crawlers explicitly and declares how Banxa prefers this
documentation to be used. Answer-engine crawlers are welcome. A
[sitemap](https://docs.banxa.com/sitemap.xml) lists every canonical page.