Powered by PST.AG

Unified MCP Gateway for Every AI Agent

Connect any LLM and any MCP client through a single, secure gateway. Route traffic to OpenAI, Anthropic, Gemini, Bedrock, and more - with built-in API key management, rate limiting, and observability.

Get an API Key → View Supported LLMs
Overview

Enterprise-Grade AI Connectivity

PST.AG's MCP Gateway gives you a single endpoint for all your AI agent traffic, backed by enterprise security, virtual API keys, and full observability.

⚙

Unified LLM Access

Route to any major LLM provider through one OpenAI-compatible endpoint. AgentGateway handles format translation internally.

🔒

Virtual API Keys

Issue scoped, revocable keys per team or application. Set token budgets, model access limits, and rate limits per key.

📈

Usage & Observability

Track token spend, latency, and error rates per key and per model. Export to OpenTelemetry, Jaeger, or your own dashboards.

🛡

MCP Protocol Support

Full support for Model Context Protocol including Streamable HTTP, SSE, and stdio transports. Federate tools across any MCP server.

🔒

A2A Agent-to-Agent

Enable secure agent-to-agent communication with capability discovery, modality negotiation, and task collaboration.

📜

Guardrails & Policies

Apply CEL-based RBAC, content filtering, prompt enrichment, and budget controls. Protect your infrastructure from misuse.

Pricing

Get Your API Key

Choose the plan that fits your usage. All plans include access to our global MCP Gateway endpoint, virtual key management, and observability dashboard.

Starter
€xx /month
For individual developers and small teams getting started with AI agent integration.
  • 1 virtual API key
  • Up to 3 LLM provider routes
  • MCP endpoint access
  • Basic usage dashboard
  • Email support
Get Started
Professional
€yy /month
For growing teams that need multiple keys and full provider coverage.
  • Up to 10 virtual API keys
  • All LLM provider routes
  • Full MCP + A2A support
  • Token budgets & rate limits per key
  • OpenTelemetry export
  • Priority email & chat support
Get Started
Enterprise
Custom
For organizations with custom requirements, dedicated infrastructure, and advanced compliance needs.
  • Unlimited virtual API keys
  • All providers + custom providers
  • Dedicated gateway instance
  • SAML/SSO integration
  • SLA guarantees
  • Dedicated support engineer
Contact Sales
1

Choose a Plan

Select the plan that matches your usage volume and feature needs.

2

Receive Your Key

We provision your API key and send credentials within 24 hours.

3

Configure Your Clients

Point your LLM clients and MCP tools to our gateway endpoint.

4

Monitor & Scale

Track usage in the dashboard, adjust budgets, and scale as needed.

LLM Providers

Connect Any LLM

AgentGateway routes requests to all major LLM providers through a single unified endpoint. Configure once, use everywhere.

OpenAI

First-class
Route to GPT-4o, GPT-4, o-series models. The gateway uses the OpenAI-compatible format internally, so any client that speaks OpenAI can reach any backend provider.
Gateway config
llm: models: - name: "*" provider: openAI params: apiKey: "$OPENAI_API_KEY"
Endpoint: https://frvat.pst.ag/v1/chat/completions

Anthropic Claude

First-class
Route to all Claude models (Opus, Sonnet, Haiku) including extended thinking, structured outputs, and token counting. The gateway handles format translation automatically.
Gateway config
llm: models: - name: "*" provider: anthropic params: apiKey: "$ANTHROPIC_API_KEY"
Endpoint: https://frvat.pst.ag/v1/messages

Google Gemini

First-class
Route to Gemini 2.5 Flash, Pro, and all Gemini models via the gateway's unified endpoint. Automatic format translation from OpenAI format.
Gateway config
llm: models: - name: "*" provider: gemini params: apiKey: "$GEMINI_API_KEY"
Endpoint: https://frvat.pst.ag/v1/chat/completions

Amazon Bedrock

First-class
Access foundation models from Amazon Bedrock including Claude on Bedrock, Amazon Nova, and Llama. Uses AWS SigV4 authentication automatically.
Gateway config
llm: models: - name: "*" provider: bedrock params: awsRegion: us-west-2
Endpoint: https://frvat.pst.ag/v1/chat/completions

Google Vertex AI

First-class
Route to models on Google Cloud Vertex AI including Gemini models via Vertex. Uses Application Default Credentials for authentication.
Gateway config
llm: models: - name: gemini-2.5-flash provider: vertex params: model: google/gemini-2.5-flash-lite-preview-06-17 vertexProject: my-project-id vertexRegion: us-west2
Endpoint: https://frvat.pst.ag/v1/chat/completions

Azure AI

First-class
Connect to models hosted on Azure AI Foundry, including OpenAI models via Azure and Claude models via Azure AI Foundry.
Gateway config
llm: models: - name: "*" provider: azure params: apiKey: "$AZURE_API_KEY" baseUrl: "https://your-resource.openai.azure.com"
Endpoint: https://frvat.pst.ag/v1/chat/completions

Ollama (Local)

First-class
Serve local models through the gateway. Point at your Ollama instance - no cloud API key needed. Perfect for development and private inference.
Gateway config
llm: port: 3000 models: - name: "*" provider: ollama params: model: llama3.2
Endpoint: https://frvat.pst.ag/v1/chat/completions

OpenAI-Compatible

Supported
Any provider that exposes an OpenAI-compatible API can be routed through the gateway: xAI (Grok), Cohere, Together AI, Groq, DeepSeek, Mistral, Perplexity, Fireworks AI, vLLM, LM Studio, and more.
Gateway config (example: xAI)
llm: models: - name: "*" provider: openAI params: apiKey: "$XAI_API_KEY" baseUrl: "https://api.x.ai/v1"
Endpoint: https://frvat.pst.ag/v1/chat/completions
MCP Clients

Connect Your Tools

Point any MCP-compatible AI client at the PST.AG MCP Gateway. Use your virtual API key for authentication. All clients connect to the same gateway endpoint.

💻 Claude Desktop

Edit ~/Library/Application Support/Claude/claude_desktop_config.json:
{ "mcpServers": { "pst-gateway": { "url": "https://frvat.pst.ag/mcp", "headers": { "Authorization": "Bearer YOUR_API_KEY" } } } }

💻 Claude Code CLI

Add via CLI or edit .mcp.json in your project:
# Via CLI claude mcp add pst-gateway \ --transport http \ https://frvat.pst.ag/mcp # Or in .mcp.json: { "mcpServers": { "pst-gateway": { "type": "http", "url": "https://frvat.pst.ag/mcp", "headers": { "Authorization": "Bearer YOUR_API_KEY" } } } }

💻 Cursor

Edit ~/.cursor/mcp.json:
{ "mcpServers": { "pst-gateway": { "url": "https://frvat.pst.ag/mcp", "headers": { "Authorization": "Bearer YOUR_API_KEY" } } } }

💻 Windsurf / Codeium

Edit ~/.codeium/windsurf/mcp_config.json:
{ "mcpServers": { "pst-gateway": { "url": "https://frvat.pst.ag/mcp", "headers": { "Authorization": "Bearer YOUR_API_KEY" } } } }

💻 VS Code / Cline

Edit ~/.config/Code/User/globalStorage/saoudrizwan.claude-dev/settings/cline_mcp_settings.json:
{ "mcpServers": { "pst-gateway": { "url": "https://frvat.pst.ag/mcp", "headers": { "Authorization": "Bearer YOUR_API_KEY" } } } }

💻 Continue.dev

Edit ~/.continue/config.json (YAML in continue.json):
{ "experimental": { "mcpServers": { "pst-gateway": { "url": "https://frvat.pst.ag/mcp", "headers": { "Authorization": "Bearer YOUR_API_KEY" } } } } }

💻 Zed

Edit ~/.config/zed/settings.json:
{ "mcp_servers": { "pst-gateway": { "url": "https://frvat.pst.ag/mcp", "headers": { "Authorization": "Bearer YOUR_API_KEY" } } } }

💻 Gemini CLI

Add via CLI or edit .mcp.json:
# Via CLI gemini mcp add pst-gateway \ --transport http \ https://frvat.pst.ag/mcp # Or in .mcp.json: { "mcpServers": { "pst-gateway": { "type": "http", "url": "https://frvat.pst.ag/mcp", "headers": { "Authorization": "Bearer YOUR_API_KEY" } } } }

💻 Goose

Edit Goose configuration YAML:
mcp_servers: pst-gateway: transport: http url: https://frvat.pst.ag/mcp headers: Authorization: "Bearer YOUR_API_KEY"

💻 Any MCP Client

Generic Streamable HTTP configuration:
{ "mcpServers": { "pst-gateway": { "type": "http", "url": "https://frvat.pst.ag/mcp", "headers": { "Authorization": "Bearer YOUR_API_KEY" } } } }

LLM Client Configuration

For connecting your coding agents' LLM calls through the gateway (not MCP), point their base URL to the gateway endpoint and use your virtual API key.

Universal OpenAI-compatible client config
BASE_URL: https://frvat.pst.ag/mcp
API_KEY: <your virtual API key>
MODEL: any model name (routed to the correct provider)