vybenetwork/solana-mcp-vybe

What is this project?

Solana MCP by Vybe is a public server that lets AI‑powered coding assistants (Cursor, Claude, Codex, etc.) talk to the Solana blockchain. It implements the Model Context Protocol (MCP), exposing Solana’s API endpoints through a single HTTP stream that the assistants can call as if they were built‑in tools.

How does it work?

MCP endpoint – All requests go to:
```
https://mcp.vybenetwork.xyz/mcp
```
The same URL is listed in the official MCP registry (server.json).
Client configuration – Different AI clients need a tiny JSON snippet:
- Cursor / Windsurf / Codex‑style – add an entry under mcpServers with type: "http" and the URL above.
- Claude Desktop – run npx mcp-remote https://mcp.vybenetwork.xyz/mcp.
- Claude Connectors Directory – just supply the name and URL.
- Claude Code / Cowork plugin – the repo contains a /.claude-plugin/plugin.json manifest that can be validated and submitted.
No secrets in the repo – All configurations point to the public endpoint; you authenticate only when the AI client prompts you.

What can the AI assistants do?

The repository defines a set of tools that the MCP server makes available:

Tool	What it does
`list‑endpoints`	Shows every Solana API path, HTTP method and a short description.
`search‑endpoints`	Finds paths, operations or schemas matching a keyword.
`get‑endpoint`	Returns the full OpenAPI description for a specific path/method.
`query‑vybe‑api`	Performs a live, authenticated GET request to any `/v4` Solana endpoint.
`query‑vybe‑api‑batch`	Same as above but can read many wallets in a single call.
`build‑vybe‑transaction`	Generates an unsigned payload for actions such as token swaps, closing accounts, or withdrawing MEV.
`pay‑with‑x402`	Gives guidance on using the x402 pay‑per‑call model for API usage.

These tools let the AI model browse the Solana API, fetch on‑chain data, and even construct transaction payloads that a user can later sign and submit.

Repository contents (as described in the README)

server.json – Metadata published to the official MCP Registry (via a GitHub Action). It contains the server’s name (io.github.vybenetwork/solana-mcp-vybe), version, and public URL.
.mcp.json / mcp.json – Root configuration files used by directory UIs (e.g., Cursor’s “Submit a Plugin”). They point to the same HTTP endpoint and contain no secrets.
.codex-plugin/plugin.json – Manifest for Codex‑style local plugins.
.claude-plugin/plugin.json – Manifest for Claude Code / Cowork plugins, describing the tools above.
claude-connector-submission.json – Ready‑to‑copy fields for submitting the connector to Claude’s directory.
chatgpt-app-submission.json – Draft fields for a future ChatGPT App listing.
examples/cursor-mcp.json – A ready‑to‑paste snippet for Cursor users.

Publishing a new version

Tag the repo (the GitHub Action reads the tag and writes it into server.json):
```
git tag v1.0.0
git push origin v1.0.0
```
Or trigger the Publish to MCP Registry workflow manually.
The workflow uses GitHub OIDC, so no personal access token is needed.

After publishing, verify the entry with:

curl -sS "https://registry.modelcontextprotocol.io/v0.1/servers?search=vybenetwork" | jq .

For organizations that enforce SSO, you may need to re‑grant OAuth access as described in the README.

Who might use this?

Developers building AI‑assisted Solana tools who want their assistants to query on‑chain data or draft transactions without writing custom API wrappers.
AI platform teams (Cursor, Claude, Codex, etc.) that need a ready‑made MCP server to add Solana support to their plugin directories.
End users of those AI assistants who can now ask questions like “What’s the balance of wallet X?” or “Create an unsigned swap transaction” and get a concrete response generated by the model.

Quick start for a user

Copy the appropriate JSON snippet from examples/ into your AI client’s configuration.
Restart the client (or reload plugins).
In the chat, invoke one of the tools, e.g., list‑endpoints or query‑vybe‑api.
Follow any authentication prompt (OAuth) that the client shows; the repo itself never stores keys.

Links mentioned in the README

MCP endpoint: https://mcp.vybenetwork.xyz/mcp
Registry metadata: https://mcp.vybenetwork.xyz
Documentation site: https://docs.vybenetwork.com/docs/mcp
Cursor Directory: https://cursor.directory
Official MCP Registry: https://registry.modelcontextprotocol.io

Bottom line

Solana MCP by Vybe is a small but functional open‑source bridge that lets modern AI coding assistants interact with Solana’s blockchain via the Model Context Protocol. It provides ready‑made configuration files for the major AI platforms, a set of useful Solana‑focused tools, and an automated publishing pipeline to keep the public server listing up‑to‑date.