Dynamic MCP Server
BYO Agent
One Connection | Infinite Possibilities
No groups to display.
MCP Endpoint
Remote MCP server (HTTPS)
https://api.agentpmt.com/mcp100% Cloud-Based Execution
The MCP server cannot edit, read, or access anything on your computer.
All tools execute on our secure cloud infrastructure
No local file system access or permissions required
The local package is just a connector that routes requests to our hosted MCP server
Local Installation
STDIO connector for Claude Code, Codex, Cursor, Zed, and other LLMs that require STDIO or custom connections
npm install -g @agentpmt/mcp-routeragentpmt-setupThis lightweight connector routes requests to https://api.agentpmt.com/mcp using your dashboard Bearer token via AGENTPMT_BEARER_TOKEN. All tool execution happens in the cloud.
1
Install for All Tools
8+
AI Platforms
5MB
Binary Size
100%
Traceable & Auditable
Included with Every Plan
The Dynamic MCP Server comes free with every AgentPMT account. You only pay for the tools you use, with full budget control and transparent pricing.
$0
Server Cost
The MCP router is completely free
Pay-per-use
Tool Pricing
Only pay for tools you actually use
Full Control
Budget Limits
Set spending caps per budget
Manage Your Entire Organization
Control agent tools across your whole team from one centralized dashboard. No technical overhead, just a few clicks.
Team-Wide Tool Approval
Approve or restrict tools for your entire organization with a single click. Changes propagate instantly to all agents.
Budget Controls
Set spending limits per team or per budget. Automatic enforcement prevents overruns.
Instant Deployment
New tools become available to all agents within 30 minutes. No reinstalls, no configuration changes.
Centralized Dashboard
Team Members24
Approved Tools48
Active Budgets12
All updates propagate in real-time
Integrated Agent Tracking and Control
Track agent activity, tool usage, and spend from one place. Set approvals and enforce budgets server-side with a complete audit trail.
Get the MCP Router on npm
Install the official package or review release notes on the npm registry page for @agentpmt/mcp-router.
Revolutionary Features
Built from the ground up to redefine how AI agents access and use tools
One Install, Unlimited Access
Install once and gain access to any tool in the marketplace. No additional installations required.
Centralized Management
Manage all agent tools for your entire organization from one dashboard with just a few clicks
Universal Binary
Single 5MB binary works across 8+ AI platforms without modification
Auto-Refresh
Tool catalog updates automatically every 30 minutes without restarts
Persistent Instances
All agent instances are persistent with full activity tracking and session continuity
Remote Execution
All tools run on remote servers - no local dependencies required
Complete Auditability
Every tool call and response is logged and traceable for compliance and debugging
Integrated Billing
Pay-per-use model with intelligent budget controls and transparent spending limits
Interactive Installer
Auto-detects AI platforms and configures them with a single command
How It Works
Simple architecture, powerful results
1
AI Assistant Requests Tools
Claude, Cursor, or any supported platform sends a JSON-RPC request over stdio
2
Router Fetches from API
The local STDIO connector calls https://api.agentpmt.com/mcp (using your dashboard Bearer token) to fetch your personalized tool catalog
3
Dynamic Tools Appear
Your approved tools with pricing and schemas are returned to the AI assistant
Tool Execution
When the AI calls a tool, the connector proxies the MCP request to https://api.agentpmt.com/mcp with server-side billing and budget enforcement
Results Returned
Tool output is streamed back to the AI assistant in real-time
AI Assistant (Claude/Cursor/etc)
↓ stdio (JSON-RPC)
agentpmt-router (local STDIO connector)
↓ HTTPS (Authorization: Bearer ...)
AgentPMT MCP endpoint (https://api.agentpmt.com/mcp)Complete Traceability & Auditability
Every action is logged, tracked, and auditable. Perfect for compliance, debugging, and organizational oversight.
Tool Call Logging
Every tool invocation is recorded with full parameters, timestamps, and user context.
Response Tracking
All tool responses, errors, and execution times are captured for full session reconstruction.
Persistent Sessions
Agent instances persist across restarts with full activity history always accessible.
Structured Audit Trail
Every tool execution is captured with complete context in an easy-to-navigate format
Instance ID:
0b2572d7-3645-4b33-b731-5b95cbdf3c74Instance Metadata:
{
"client": "claude_code",
"platform": "linux",
"client_info": {
"ip_address": "69.136.156.79",
"user_agent": "AgentPMT-MCP/1.0"
}
}Request Payload:
{
"product_id": "692720cb46dac36e67d0ef2e",
"parameters": {
"action": "calculate",
"expression": "integrate(x^2, x)"
},
"budget_id": "695090103119f650009bc8893"
}Response Data:
{
"success": true,
"instance_id": "0b2572d7-3645-4b33-b731-5b95cbdf3c74",
"response": {
"status_code": 200,
"data": {
"success": true,
"output": {
"result": "x^3/3 + C",
"execution_time": "0.23s"
}
}
},
"cost": "$0.06",
"timestamp": "2026-01-03T14:32:16Z"
}Complete Context
Every detail captured and organized
Easy Navigation
Structured format for quick review
Full Traceability
From request to response
Works with Any MCP-Compatible Agent
The Dynamic MCP Server connects to any AI agent that supports the Model Context Protocol. Our easy setup tool automatically configures these popular platforms:
Claude Desktop
Claude Code CLI
Cursor
VS Code
Windsurf
Zed
OpenAI Codex CLI
Google Gemini CLI
Universal Configuration
{
"command": "npx",
"args": ["--package=@agentpmt/mcp-router@latest", "agentpmt-router"],
"env": {
"AGENTPMT_BEARER_TOKEN": "your-dashboard-bearer-token"
}
}This configuration works with any MCP-compatible agent. Our installer automates the setup for popular platforms.
AgentPMT Dynamic MCP Features
See how we compare to standard MCP servers
| Feature | Standard MCP Server | AgentPMT Dynamic MCP |
|---|---|---|
| Tool Access | Limited, pre-installed | Unlimited marketplace access |
| Tool Loading | Hardcoded in source | Dynamic from API |
| Updates | Requires rebuild | Automatic |
| Execution | Local process | Remote API |
| Organization Management | Individual config | Centralized dashboard |
| Auditability | Limited or none | Full activity logging |
Technical Innovations
Enterprise-grade features built for production
Runtime Tool Loading
Tools are fetched from the API on each tools/list request. No recompilation needed - new tools appear automatically when you approve them on your dashboard or when tools are added to the marketplace.
Automatic Change Detection
The server polls the API every 30 minutes and sends MCP notifications when your tool catalog changes. The AI assistant automatically refreshes its tool list.
Integrated Billing & Budget Control
Every tool call includes Bearer-token authentication and server-side budget enforcement. Users set spending limits in the dashboard and the server enforces them with full audit trails.
Persistent & Auditable Sessions
All agent instances are persistent with complete activity logging. Every tool call, response, and error is recorded for compliance, debugging, and organizational oversight.
Control Your Agents On The Go
The AgentPMT mobile app extends your Dynamic MCP setup to your phone. Get notified the moment an agent needs a decision and respond in seconds.
Real-Time Notifications
Push alerts the moment an agent needs approval -- tool access, workflows, or payments.
One-Tap Approvals
Review request details and approve or deny without opening a laptop.
Secure Payments
Biometric authentication for every credit card authorization. Card data is tokenized and encrypted.
Full Activity Feed
See every tool call, response, and status update across all your budgets in real time.