MCP-Inscription Server

The MCP-Inscription Server lets AI models read content from Ordinals Inscriptions embedded in Bitcoin transactions. It detects inscriptions, decodes their data, and exposes it for use by client applications and AI agents, enabling richer interactions with ordinal content.

How to use

You can use the MCP-Inscription server by connecting a client that supports the Model Context Protocol (MCP). There are two ways to integrate, depending on your setup: run the server locally as a standard I/O extension for your agent framework, or run it as a remote service and connect via Server-Sent Events. In either case, you’ll be able to request decoded inscription data by supplying a transaction identifier and then receive the decoded content such as text, JSON, or images.

How to install

Prerequisites: you need Node.js installed on your machine. If you plan to build from source, you’ll also need npm.

Step-by-step install and run flow as described here involves cloning the project, installing dependencies, building, and then starting the CLI server. Use the concrete commands below.

Configuration and run options

There are two primary ways to run the MCP-Inscription server and connect with Goose or Claude Desktop: as a local STDIO extension or as a remote SSE service. The CLI exposes both modes through a built script.

Security and maintenance notes

Keep your server and dependencies up to date. If you expose the server over the network, ensure you manage access and monitor logs for errors or unusual activity.

Testing integration scenarios

Test interactions by starting a Goose session with the local STDIO extension or by querying the remote SSE endpoint. Validate that decoded inscription data is returned correctly and that the server can handle typical inscription formats (text, JSON, images, etc.).

Project structure overview

The project organizes code into source, server transports, utilities, and tooling. Core functionality includes ordinals parsing, SSE/stdio server implementations, and a CLI launcher.

Troubleshooting quick starts

If the extension doesn’t appear in Goose or Claude Desktop, verify the server process is running, the correct path to the CLI script is provided, and that the extension is enabled in your client’s settings.

Available tools

show_ordinals

Decodes Ordinal inscription data from a transaction's witness data, returning the decoded content which may be text, JSON, HTML, or other formats.