About Delora MCP
Delora MCP connects MCP clients to the Delora API for cross-chain quotes and reference data. It provides tools for querying supported chains, tokens, and available tools.
Key features
- 01Fetch cross-chain quotes
- 02List supported chains
- 03List supported tokens
- 04Look up a token by chain
- 05Retrieve available tools
Use cases
- 01Get a cross-chain quote from an MCP client
- 02Check which chains and tokens Delora supports
MCP support
Exposes Delora API queries as MCP tools, including quotes, chains, tokens, and available tools. Supports stdio and streamable HTTP transports.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://mcp.delora.build/mcpclaude mcp add --transport http deloraprotocol-delora https://mcp.delora.build/mcp{"mcpServers":{"deloraprotocol-delora":{"type":"http","url":"https://mcp.delora.build/mcp"}}}Open in Cursor{"servers":{"deloraprotocol-delora":{"type":"http","url":"https://mcp.delora.build/mcp"}}}Open in VS Code{"mcpServers":{"deloraprotocol-delora":{"type":"http","url":"https://mcp.delora.build/mcp"}}}{"mcpServers":{"deloraprotocol-delora":{"serverUrl":"https://mcp.delora.build/mcp"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
https://mcp.delora.build/mcpAuth requiredTools
6This tool takes no parameters.
Details
From the project READMEDelora MCP
MCP (Model Context Protocol) server for the Delora API: cross-chain quotes, supported chains, tokens, and tools.
Install
npm install
Configuration
Copy env.example to .env and adjust:
| Variable | Description | Default |
|---|---|---|
DELORA_API_URL | Delora API base URL | https://api.delora.build |
DELORA_API_KEY | Optional API key forwarded to Delora API as x-api-key | unset |
MCP_TRANSPORT | stdio or http | stdio |
PORT | HTTP server port (when MCP_TRANSPORT=http) | 3000 |
HOST | HTTP bind address | 0.0.0.0 |
When DELORA_API_KEY is set, the MCP server forwards it to Delora API as the x-api-key header. In HTTP mode, the server also accepts incoming x-api-key or Authorization: Bearer ... headers and forwards the resolved key upstream. For HTTP requests, incoming headers take priority over DELORA_API_KEY. For stdio, DELORA_API_KEY remains the way to provide the key.
Run
Stdio (default) — for Cursor/IDE MCP over stdio:
npm run dev
# or
npm run build && npm start
HTTP — for streamable HTTP (e.g. behind a reverse proxy):
npm run dev:http
# or
npm run build && npm run start:http
Server listens on http://0.0.0.0:3000/mcp.
Cursor configuration
Stdio:
{
"mcpServers": {
"delora": {
"command": "node",
"args": ["/path/to/delora-mcp/dist/index.js"]
}
}
}
Streamable HTTP (e.g. after deploying to https://mcp.delora.build):
{
"mcpServers": {
"delora": {
"url": "https://mcp.delora.build/mcp",
"transport": "streamable-http",
"headers": {
"x-api-key": "YOUR_API_KEY"
}
}
}
}
You can also use Authorization: Bearer YOUR_API_KEY instead of x-api-key when your MCP client supports custom HTTP headers.
Docker
Build and run:
docker build -t delora-mcp .
docker run -p 3000:3000 -e MCP_TRANSPORT=http delora-mcp
For production behind a domain (e.g. mcp.delora.build): run the container with MCP_TRANSPORT=http and put nginx/traefik (or another reverse proxy) in front with HTTPS; then use the URL https://mcp.delora.build/mcp with streamable-http in Cursor.
API
- Delora API — base URL for all
/v1/*endpoints (quotes, chains, tokens, tools, token).
Tools
| Tool | Description |
|---|---|
get_instructions | Returns this guide (call first for workflow). |
get_quote | GET /v1/quotes — cross-chain quote. |
get_chains | GET /v1/chains — supported chains. |
get_tools | GET /v1/tools — available tools. |
get_tokens | GET /v1/tokens — supported tokens. |
get_token | GET /v1/token — single token by chain + token. |
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.