io.github.debridge-finance/debridge-mcp MCP Server
io.github.debridge-finance/debridge-mcp
Cross-chain and same-chain swaps, fee estimation, and trade execution via AI agents.
What is the io.github.debridge-finance/debridge-mcp MCP server?
The deBridge MCP server integrates deBridge's cross-chain and same-chain swap capabilities into AI agents like Claude and Cursor. It provides tools for token search, chain support discovery, transaction creation, and trade execution across major blockchain networks.
deBridge MCP enables AI agents to execute swaps and estimate fees across multiple blockchain networks. Use it to automate cross-chain trading, discover supported tokens and chains, and execute transactions programmatically without manual intervention.
How to install io.github.debridge-finance/debridge-mcp
Copy-paste configuration for popular MCP clients.
Tools & capabilities
Tools this server exposes to the agent.
get_instructions— Retrieve instructions for using deBridge MCP tools and workflows.search_tokens— Search for supported tokens across blockchain networks.get_supported_chains— List all blockchain networks supported by deBridge.create_tx— Create a transaction for cross-chain or same-chain swaps.transaction_same_chain_swap— Execute a same-chain swap transaction.
Use cases
- Execute cross-chain token swaps programmatically via AI agents
- Estimate swap fees and get quotes before executing trades
- Search for supported tokens and discover available blockchain networks
- Automate trading workflows across multiple chains without manual interaction
- Build AI-powered trading bots that interact with deBridge liquidity
io.github.debridge-finance/debridge-mcp MCP server FAQ
deBridge MCP is a Model Context Protocol server that gives AI agents access to deBridge's cross-chain and same-chain swap infrastructure, including token search, fee estimation, and trade execution.
The MCP server itself is free and open-source (MIT licensed). Swap execution costs depend on deBridge's protocol fees and blockchain gas costs.
Add to `.cursor/mcp.json`: `{"mcpServers": {"debridge": {"url": "https://agents.debridge.com/mcp"}}}` for the hosted endpoint, or use the npx command for a local proxy.
Use `claude mcp add debridge --transport http https://agents.debridge.com/mcp` for the CLI, or add the remote URL in Claude Web/Desktop apps.
The hosted endpoint at agents.debridge.com/mcp does not require API keys. Actual swap execution may require wallet connection depending on your agent framework.
Yes, use `npx @debridge-finance/debridge-mcp@latest` for stdio mode or Docker for HTTP proxy mode. The local process proxies requests to the hosted endpoint.
README (reference)
Source of truth, from the repository.
deBridge MCP
deBridge MCP integration for AI agents — cross-chain and same-chain swaps, fee estimation, and trade execution across major blockchain networks.
The hosted MCP endpoint is available at agents.debridge.com. This npm package is a thin proxy client for agents that require a local stdio or HTTP process.
https://github.com/user-attachments/assets/8ebe88ff-db3c-455e-9efb-50389e4bf5bd
Connecting to deBridge MCP
Recommended: Direct connection (Streamable HTTP)
Connect your agent directly to the hosted endpoint — no installation, no local process:
https://agents.debridge.com/mcp
This is a Streamable HTTP MCP endpoint. It exposes tools (get_instructions, search_tokens, get_supported_chains, create_tx, transaction_same_chain_swap), workflow skills via MCP resources, and an llms.txt overview. Tools and skills are maintained in the debridge-finance/debridge-skills repository.
Generic Streamable HTTP configuration:
"debridge": {
"type": "streamable-http",
"url": "https://agents.debridge.com/mcp"
}
<details>
<summary>Claude Code (CLI & IDE plugins)</summary>
claude mcp add debridge --transport http https://agents.debridge.com/mcp
Verify:
claude mcp list
</details>
<details>
<summary>Claude Web & Desktop Apps</summary>
Add as a remote MCP server with the URL https://agents.debridge.com/mcp.
Add to .cursor/mcp.json in your project or ~/.cursor/mcp.json globally:
{
"mcpServers": {
"debridge": {
"url": "https://agents.debridge.com/mcp"
}
}
}
</details>
<details>
<summary>GitHub Copilot (VS Code Chat)</summary>
Add to .vscode/mcp.json in your workspace:
{
"mcpServers": {
"debridge": {
"type": "streamable-http",
"url": "https://agents.debridge.com/mcp"
}
}
}
</details>
<details>
<summary>Windsurf</summary>
Add to ~/.codeium/windsurf/mcp_config.json:
{
"mcpServers": {
"debridge": {
"type": "streamable-http",
"url": "https://agents.debridge.com/mcp"
}
}
}
</details>
<details>
<summary>Cline</summary>
Open Cline settings in VS Code, go to MCP Servers, click "Edit MCP Settings" and add:
{
"mcpServers": {
"debridge": {
"type": "streamable-http",
"url": "https://agents.debridge.com/mcp"
}
}
}
</details>
<details>
<summary>Continue</summary>
Add to ~/.continue/config.json:
{
"mcpServers": [
{
"name": "debridge",
"transport": "streamable-http",
"url": "https://agents.debridge.com/mcp"
}
]
}
</details>
<details>
<summary>Zed</summary>
Add to your Zed settings (~/.config/zed/settings.json):
{
"context_servers": {
"debridge": {
"transport": "streamable-http",
"url": "https://agents.debridge.com/mcp"
}
}
}
</details>
<details>
<summary>OpenClaw (via mcp-adapter plugin)</summary>
Install the mcp-adapter plugin:
openclaw plugins install mcp-adapter
Add to ~/.openclaw/openclaw.json:
{
"plugins": {
"entries": {
"mcp-adapter": {
"enabled": true,
"config": {
"servers": [
{
"name": "debridge",
"transport": "streamable-http",
"url": "https://agents.debridge.com/mcp"
}
]
}
}
}
}
}
</details>
Alternative: Run a local proxy
Some agent frameworks only support stdio transport and cannot connect to a remote HTTP endpoint directly. This npm package bridges that gap — it runs a local MCP process that transparently proxies all requests to https://agents.debridge.com/mcp.
Tools and resources are not implemented locally. They are dynamically discovered from the upstream endpoint at startup (stdio) or forwarded per-request (HTTP), so any changes to the hosted MCPd are reflected automatically without updating the package.
How the proxy works
-
Stdio mode (default): Opens a long-lived MCP client connection to the upstream MCPd via
StreamableHTTPClientTransport. Creates a localServerwithStdioServerTransportthat mirrors the upstream capabilities and forwardstools/list,tools/call,resources/list,resources/read, andresources/templates/listto the upstream. -
HTTP mode (
MCP_TRANSPORT=http): Transparent HTTP reverse proxy via Express. Forwards raw JSON-RPC requests and SSE response streams to the upstream MCPd, passing throughMcp-Session-Idheaders for session continuity.
npx (stdio)
npx -y @debridge-finance/debridge-mcp@latest
stdio configuration:
"debridge": {
"type": "stdio",
"command": "npx",
"args": ["-y", "@debridge-finance/debridge-mcp@latest"]
}
<details>
<summary>Claude Code (CLI & IDE plugins)</summary>
claude mcp add debridge npx -- -y @debridge-finance/debridge-mcp@latest
</details>
<details>
<summary>Cursor</summary>
OR install from Cursor Directory
OR add to .cursor/mcp.json:
{
"mcpServers": {
"debridge": {
"command": "npx",
"args": ["-y", "@debridge-finance/debridge-mcp@latest"]
}
}
}
</details>
Docker (HTTP proxy)
docker build -t debridge-mcp .
docker run -p 3000:3000 debridge-mcp
Then connect your agent to http://localhost:3000/mcp using the Streamable HTTP configs above.
Environment Variables
| Variable | Default | Description |
|---|---|---|
REMOTE_MCP_URL | https://agents.debridge.com/mcp | Remote MCP endpoint to proxy to |
MCP_TRANSPORT | stdio | Transport mode: stdio or http |
PORT | 3000 | HTTP listen port (HTTP mode only) |
HOST | 0.0.0.0 | HTTP listen host (HTTP mode only) |
Development
# Install dependencies
npm install
# Run in dev mode (stdio proxy)
npm run dev
# Run in dev mode (HTTP proxy)
npm run dev:http
# Test with MCP Inspector
npm run inspect
# Run tests
npm test
Links
License
MIT, Copyright 2026 deBridge
Related MCP servers

io.github.debu-sinha/excalidraw
Security-hardened Excalidraw MCP server with auth, rate limiting, and 14 tools

Stable eight-tool MCP server for changed-UI assurance, authority-aware UI task context, and repair.

Give your AI agents the ability to listen. Microphone capture and speech-to-text.

Decide Policy Notaries
Fail-closed refund, cancellation, return, and trial checks for 100 US subscription vendors.

Google Calendar access for LLM agents: find, create, update, delete events with scheduling intelligence.
MCP server for Deckbase - exposes docs, template blocks, and project context
View repository →