Install AgentPMT MCP Server for Openclaw

A specialized skill to connect MCP-compatible AI agents, IDEs, and programmatic clients to the AgentPMT tool marketplace with budget enforcement.

agentpmt
v1.0.0
Jun 10, 2026
0
451
0

Install & Download

1. ClawHub CLI

The fastest way to install a skill directly from the registry.

npx clawhub@latest install install-agentpmt-mcp

2. Manual Installation

Copy the skill folder to one of these locations

Global
~/.openclaw/skills/
Workspace
<project>/skills/

Priority: Workspace > Local > Bundled

3. Prompt Installation

Copy this prompt to OpenClaw to install it automatically.

Help me install install-agentpmt-mcp using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).

Prefer to download?

Get the raw skill files in a ZIP archive.

What is Install AgentPMT MCP Server?

The AgentPMT MCP Server skill allows developers and teams to seamlessly connect any Model Context Protocol (MCP) compatible AI agent to the AgentPMT tool marketplace. By installing this capability through Openclaw Skills, you can grant your local or cloud-based agents access to a vast, dynamic catalog of commercial APIs, advanced tools, and paid developer services. This system bridges the gap between local execution environments and premium cloud integrations without compromising security or budget controls.

With this integration, all interactions are secured and constrained by strict spending caps and budgets that you define. The skill utilizes a lightweight local STDIO connector or direct HTTPS connections to dynamically present approved tool sets to platforms like Claude Desktop, Cursor, VS Code, and programmatic agents. Leveraging Openclaw Skills simplifies this deployment, ensuring your AI agents are always equipped with the right capabilities while preventing run-away spending.

Install AgentPMT MCP Server Use Cases

  • Seamlessly connect local desktop environments (such as Claude Desktop, Claude Code, Windsurf, or Cursor) to external paid APIs and marketplaces.
  • Enforce strict programmatic budget constraints and financial spending limits on AI-driven tasks and agent runtimes.
  • Deploy remote, programmatic agents connected to the AgentPMT cloud using direct, secure HTTPS streaming JSON-RPC endpoints.
  • Access a single, managed hub of diverse tools and services without having to manage separate individual API keys or billing accounts for each vendor.

How Install AgentPMT MCP Server Works

  1. Authentication Verification: The server extracts and decodes a combined base64-encoded bearer token consisting of the user's AgentPMT API Key and Budget Key.
  2. Dynamic Tool Discovery: The agent sends an initialization request to list available capabilities. AgentPMT evaluates the specific budget restrictions and returns a customized tool catalog dynamically.
  3. Local or Cloud Routing: For desktop IDEs, a lightweight @agentpmt/mcp-router acts as an STDIO relay, forwarding requests to the AgentPMT cloud. For remote agents, calls are sent directly over streamable HTTPS.
  4. Execution and Audit: Tool executions occur securely within the AgentPMT cloud infrastructure, tracking usage, deducting costs, and logging detailed audit logs in real-time.

Install AgentPMT MCP Server Setup

Prerequisites

You must have an active AgentPMT account. Prepare your credentials:

  1. API Key: Generate from your Dashboard > Account > API Keys.
  2. Budget Key: Generate from Dashboard > Budgets > [Select Budget] > Budget Keys.

Run the interactive setup CLI tool to automatically configure your active IDEs (Claude Desktop, Cursor, Claude Code, etc.):

npm install -g @agentpmt/mcp-router
agentpmt-setup

Method 2: Manual Setup via STDIO

Step 1: Generate the Bearer Token

Combine your credentials into YOUR_API_KEY:YOUR_BUDGET_KEY and base64-encode them:

echo -n "YOUR_API_KEY:YOUR_BUDGET_KEY" | base64

Step 2: Configure Your Client

Claude Desktop Edit your configuration file (~/.config/Claude/claude_desktop_config.json on macOS or %APPDATA%\\Claude\\claude_desktop_config.json on Windows):

{
  "mcpServers": {
    "agentpmt": {
      "command": "npx",
      "args": ["--package=@agentpmt/mcp-router@latest", "agentpmt-router"],
      "env": {
        "AGENTPMT_BEARER_TOKEN": "<your-base64-token>"
      }
    }
  }
}

Claude Code Add to your global or project .mcp.json / ~/.claude/mcp.json file:

{
  "mcpServers": {
    "agentpmt": {
      "command": "npx",
      "args": ["--package=@agentpmt/mcp-router@latest", "agentpmt-router"],
      "env": {
        "AGENTPMT_BEARER_TOKEN": "<your-base64-token>"
      }
    }
  }
}

Method 3: Direct HTTPS (For Programmatic Agents)

Connect directly using the endpoint https://api.agentpmt.com/mcp using standard JSON-RPC protocol over HTTP.

curl -X POST https://api.agentpmt.com/mcp \
  -H "Authorization: Bearer <your-base64-token>" \
  -H "Content-Type: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "id": 1,
    "method": "initialize",
    "params": {
      "protocolVersion": "2025-03-26",
      "clientInfo": { "name": "my-agent", "version": "1.0" },
      "capabilities": {}
    }
  }'

Install AgentPMT MCP Server Data Schema & Taxonomy

Configuration Schema

Regardless of your integration environment, the AgentPMT MCP connection uses a structured config schema. For desktop STDIO clients, write properties as environment variables:

Parameter Description Required
AGENTPMT_BEARER_TOKEN Base64-encoded string combining your API Key and Budget Key (format: api_key:budget_key). Yes

Endpoint Schemas

When utilizing Openclaw Skills to manage programmatic integrations via Direct HTTPS, requests must implement the MCP 2.0 protocol over streamable HTTP:

Method Description Content Type
initialize Verifies protocol capabilities and establishes connection metadata. application/json
tools/list Discovers authorized APIs permitted by the defined Budget Key. application/json
tools/call Executes a specific API action via AgentPMT's cloud environment. application/json

Install AgentPMT MCP Server Advanced Features

  • Dynamic Permissions: Auto-filtering of accessible tools depending on real-time budget approvals and balance limitations.
  • Pricing Metadata: Transmits execution costs and fees transparently to the LLM agent prior to tool calls, optimizing token usage and expenditures.
  • Self-Healing Diagnostics: Native internal tools like AgentPMT-Refresh-Tools and AgentPMT-Report-Tool-Issue let the AI diagnose connection issues directly.
  • Non-Invasive STDIO Router: The @agentpmt/mcp-router runs purely in-memory as a relay, protecting local workspace files from unauthorized cloud uploads.

SKILL.md


Loading

Related Openclaw Skills

METADATA

Github Stars: 0
forks: 0

Featured*