Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.delegare.dev/llms.txt

Use this file to discover all available pages before exploring further.

Delegare provides an official MCP Server that enables LLMs (like Claude or GPT-4) to perform economic actions autonomously using Intent Mandates. The server is available in the packages/mcp-server directory of the Vault repository.

​
Available Tools

The MCP server exposes the following tools to the agent:

​
1. setup_spending_mandate

Generates a setup URL for the human user.
  • Input: maxAmountPerTxCents, maxMonthlySpendCents, railPreference
  • Output: A markdown link for the user to authorize.

​
2. check_mandate_balance

Allows the agent to see how much it can spend.
  • Input: mandateId
  • Output: Remaining monthly balance.

​
3. authorize_agent_payment

The core tool for making purchases.
  • Input: amountCents, currency, description, recipient
  • Output: A cryptographic Intent Mandate (SD-JWT-VC) or a success receipt if the server handles the execution.

​
4. delegare_fetch

A tool for scraping or fetching data from third-party APIs that may be monetized.
  • Input: url, method, body, bundleToken (optional)
  • Output: The fetched HTTP response content.
  • x402 Auto-Payment: If the target URL returns a 402 Payment Required challenge, this tool intercepts it. It evaluates the payment options (Crypto USDC vs. Fiat Credit Bundle), resolves the payment using the agent’s mandate or bundle token, and resubmits the request to fetch the data.

​
5. purchase_bundle

Allows the agent to purchase an API credit bundle if a merchant requires fiat payment or if the agent doesn’t have a crypto wallet.
  • Input: merchantBundleUrl
  • Output: A checkout link for the user to complete the fiat payment via Stripe, plus available tiers.

​
6. check_bundle_balance

Checks the remaining requests on a purchased API credit bundle.
  • Input: merchantBalanceUrl, bundleToken
  • Output: Remaining credits and tier details.

​
MCP Authentication Guard

To prevent unauthorized tool calls, the MCP server implements an mcpAuthGuard.
  • Every tool call must be authorized by an OAuth session.
  • Credentials (merchantId, apiKey) are never exposed to the LLM prompt.
  • The guard injects the required metadata into the tool context based on the bearer token in the request header.

​
Running Locally with Python

The Delegare MCP Server is also published to PyPI, allowing you to run it locally or embed it within your own orchestration environments. 
pip install delegare-mcp
To run the MCP server locally over stdio: 
export DELEGARE_MERCHANT_ID="your_merchant_id"
export DELEGARE_API_KEY="your_api_key"

python -m delegare_mcp.server

​
Claude Desktop Local Config

If you prefer to run the server locally rather than connecting to the remote endpoint, configure Claude Desktop to spawn the Python process: 
{
  "mcpServers": {
    "delegare": {
      "command": "python",
      "args": ["-m", "delegare_mcp.server"],
      "env": {
        "DELEGARE_MERCHANT_ID": "your_merchant_id",
        "DELEGARE_API_KEY": "your_api_key"
      }
    }
  }
}

​
Usage with Claude.ai

Connect Delegare as a remote MCP connector on Claude.ai:
  1. Go to Settings → Connectors → Add custom connector.
  2. Set the Name to Delegare.
  3. Set the Remote MCP server URL to:
EnvironmentURL
Sandboxhttps://api.sandbox.delegare.dev/mcp
Productionhttps://api.delegare.dev/mcp
  1. Click Add. Claude.ai will automatically handle OAuth authentication when you first use a Delegare tool.

​
Usage with ChatGPT

  1. Go to Settings → Apps and enable Developer mode.
  2. Click Add app and paste the MCP URL from the table above.
  3. Open a new chat, click +, select Delegare, and start a conversation.

​
Usage in Claude Desktop

Add Delegare as a remote MCP server in your claude_desktop_config.json: 
{
  "mcpServers": {
    "delegare": {
      "type": "streamableHttp",
      "url": "https://api.sandbox.delegare.dev/mcp"
    }
  }
}