MCP Server
MCP Server Overview
A hosted Model Context Protocol server that lets AI agents send and manage signed documents on SignatureAPI.
Model Context Protocol (MCP) is an open standard that lets AI applications expose tools and data to LLMs through a uniform interface. SignatureAPI provides a hosted MCP server so any MCP-compatible agent can send envelopes, manage signing, and look up documentation on behalf of an account.
The SignatureAPI MCP server is currently in private beta. To have it enabled on your account, contact support. If you run into issues or want to share feedback, please get in touch as well.
What you can do
Through the MCP server, an agent can:
- Create envelopes with documents and recipients.
- Look up, list, cancel, and delete envelopes.
- Upload local files for use as envelope documents.
- Search the SignatureAPI documentation for context before acting.
See Tools and resources for the complete list.
Server endpoint
https://mcp.signatureapi.com/mcpThe server uses the Streamable HTTP transport defined by the MCP specification. Most clients only need this URL; they discover the rest at connection time.
Authentication model
The MCP server authenticates every request with an OAuth 2.1 access token issued through Dynamic Client Registration. Interactive clients like Claude, ChatGPT, and Cursor run the flow automatically; each user signs in once and the client manages tokens from there.
See Authorization for the flow, the discovery endpoints, and patterns for service providers.
Account scope
Every request is scoped to a single SignatureAPI account, derived from the active organization on the access token. A user who belongs to multiple accounts must authorize one at a time.
Disconnection is enforced server-side on every request, so revoking access from the dashboard takes effect immediately. See Managing connections.
When to use MCP vs. the REST API
| Use MCP when | Use the REST API when |
|---|---|
| You are integrating an AI agent (Claude, ChatGPT, Cursor, a custom agent). | You are building a traditional backend or product integration. |
| You want the agent to discover capabilities and schemas at runtime. | You want full coverage of the API surface, including resources not yet exposed via MCP. |
| You want capability discovery and tool/resource semantics tailored to LLMs. | You don’t need tool/resource semantics. |
The two surfaces are complementary. The MCP server itself is a thin layer that delegates to the REST API, so anything an agent can do through MCP, your backend can also do directly against REST.