AI and MCP

Connect Claude, Cursor, Windsurf, or any MCP client to the Verdocs docs so your coding agent can search and read them directly. Plain HTTP endpoints too.

Verdocs is an eSignature platform you build into your own product: a REST API, SDKs for the browser and the server, and embeddable web components that put signing inside your application instead of redirecting people to somebody else's portal.

If you are writing that integration with an AI assistant, point it here first. This site runs an MCP server that lets your agent search and read the documentation directly, so it works from what Verdocs actually does today rather than from whatever it remembers about eSignature APIs in general.

The two MCP surfaces

Surface	Endpoint	What it does	Status
Documentation MCP	`https://developers.verdocs.com/mcp`	Search and read these docs. No account or key needed.	Available now
Platform MCP	To be announced	Act on the Verdocs API: templates, envelopes, recipients, organization settings.	In development

The documentation MCP is read-only and anonymous. It reads a public site, so there is nothing to authenticate and nothing it can change.

Connect your client

Claude Code

claude mcp add --transport http verdocs-docs https://developers.verdocs.com/mcp

Claude Desktop, Cursor, Windsurf, and other clients

Anything that supports remote MCP over HTTP takes the URL directly. In a JSON config file that usually looks like this:

{
  "mcpServers": {
    "verdocs-docs": {
      "type": "http",
      "url": "https://developers.verdocs.com/mcp"
    }
  }
}

The exact key names move around between releases, so check your client's current docs if it rejects the block above. Clients that still speak only stdio can bridge to it:

{
  "mcpServers": {
    "verdocs-docs": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://developers.verdocs.com/mcp"]
    }
  }
}

Once it is connected, ask your assistant something like "using the Verdocs docs, show me how to send an envelope from a template" and it will search and read the relevant pages before answering.

Tools

Tool	Arguments	Returns
`search_docs`	`query` (string, up to 500 characters)	Ranked section matches, each with its breadcrumb trail, page URL, and raw markdown URL. This is the one to reach for first.
`get_doc_page`	`path` (string, up to 256 characters)	One page as markdown. Accepts `learn/mcp`, `/docs/learn/mcp`, or `/docs/learn/mcp.mdx`; all three normalize to the same page.
`list_docs_paths`	none	Every documentation path with its URL. Useful for seeing the shape of the docs when no search term is obvious.

Search runs against the same index the site's own search box uses, so an agent and a person asking the same question get the same results.

Plain HTTP, without MCP

Everything the MCP server reads is a public URL, so a RAG pipeline or a one-off script does not need MCP at all.

Any page as markdown. Append .mdx to any docs URL:

https://developers.verdocs.com/docs/guides/getting-started.mdx

This is worth knowing even if you are reading pages by hand. A rendered docs page is around 300 KB of HTML, most of it navigation and syntax-highlighting markup; the same page as markdown is a few kilobytes of content. Every page also advertises this with a <link rel="alternate" type="text/markdown"> tag, so a crawler can find it without being told.

An index of every page, with titles and descriptions, plus the API base URL and auth model:

https://developers.verdocs.com/llms.txt

The full text of the entire site in one response, for ingesting the whole knowledge base at once:

https://developers.verdocs.com/llms-full.txt

The OpenAPI specification, if you would rather generate a client than read prose:

https://developers.verdocs.com/openapi.json

Checking the connection

If a client will not connect, test the transport directly. A successful initialize returns a JSON-RPC result carrying the server's name, version, and capabilities:

curl -sS -X POST "https://developers.verdocs.com/mcp" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -H "Mcp-Protocol-Version: 2024-11-05" \
  -d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"curl","version":"1"}}}'

MCP Inspector pointed at the same URL gives you a UI for listing and calling the tools.

The endpoint speaks streamable HTTP and is stateless, so there is no session to establish and no order you have to call things in. Request bodies over 2 MiB are rejected.

Platform MCP

The documentation MCP tells an agent how Verdocs works. The platform MCP, which is in development now, will let it use Verdocs: creating and sending envelopes, checking signing status, managing templates and recipients, and reading organization configuration, all under your own API credentials and scoped to the same permissions those credentials already carry.

It will be a separate endpoint with its own authentication, so connecting to the docs today does not give an agent any access to your account, and it never will.

Watch What's new for the announcement, or subscribe to the changelog feed.

AI and MCP

On this page