pan-mcp-relay
Security-enhanced MCP Relay/Proxy server providing real-time AI threat protection for MCP Clients, scanning tool descriptions, call parameters, and responses for threats like prompt injections and malicious URLs.
README
Prisma AIRS MCP Security Relay
By Palo Alto Networks
<!--TOC-->
<!--TOC-->
<a id="overview" href="#overview">
Overview
</a>
pan-mcp-relay is a security-enhanced Model Context Protocol
(MCP) Relay (Proxy) Server providing real-time AI threat protection for MCP Clients, built with the Prisma AIRS AI Runtime API Intercept.
pan-mcp-relay will help protect MCP Clients such as IDE's, LLM Chat Clients, and AI Agents from harmful MCP Server Tools by automatically scanning and blocking
various threats, including prompt injections, malicious URLs, insecure outputs, AI agentic threats, sensitive data loss, and more.
The MCP Relay scans all MCP Server tool descriptions, tool call parameters, and tool call responses.
For licensing, onboarding, activation, and to obtain an API authentication key and profile name, refer to the Prisma AIRS AI Runtime: API Intercept Overview administration documentation.
Prerequisites
Follow the following Activation and Onboarding guides:
- Create and associate a deployment profile for Prisma AIRS AI Runtime API intercept in your Customer Support Portal.
- Onboard Prisma AIRS AI Runtime API intercept in Strata Cloud Manager.
- Manage applications, API keys, security profiles, and custom topics in Strata Cloud Manager.
Requirements for Prisma AIRS API Usage
- API Key: Generate an API Key during the onboarding process in Strata Cloud Manager (see the onboarding prerequisite step above).
- Specify the API Key via configuration flag, environment variable or configuration file.
- AI Security Profile Name: This is the API security profile you created during the onboarding process in Strata Cloud Manager (see the prerequisite step on creating an API security profile above).
- Specify the profile name or the profile ID via configuration flag, environment variable, or configuration file.
[!NOTE] You can manage API keys and AI security profiles in Strata Cloud Manager.
- Log in to Strata Cloud Manager.
- Navigate to Insights > Prisma AIRS > Prisma AIRS AI Runtime: API Intercept.
- In the top right corner, click:
- Manage > API Keys to copy, regenerate, or rotate the API key token.
- Manage > Security Profiles to fetch details or update AI security profiles.
- Manage > Custom Topics create or update custom topics for custom topic guardrails threat detections.
For complete details, refer to the Activation and Onboarding guides on how to Manage Applications, API Keys, Security Profiles, and Custom Topics.
<a id="installation" href="#installation">
Installation
</a>
We highly recommend using uv over pip or pipx. Try it! You'll love it.
<details><summary>Installing uv</summary>
See more installation methods at https://docs.astral.sh/uv/getting-started/installation/.
Standalone installer
uv provides a standalone installer to download and install uv:
=== "macOS and Linux"
Use `curl` to download the script and execute it with `sh`:
```console
$ curl -LsSf https://astral.sh/uv/install.sh | sh
```
If your system doesn't have `curl`, you can use `wget`:
```console
$ wget -qO- https://astral.sh/uv/install.sh | sh
```
Request a specific version by including it in the URL:
```console
$ curl -LsSf https://astral.sh/uv/0.8.13/install.sh | sh
```
=== "Windows"
Use `irm` to download the script and execute it with `iex`:
```pwsh-session
PS> powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
```
Changing the [execution policy](https://learn.microsoft.com/en-us/powershell/module/microsoft.powershell.core/about/about_execution_policies?view=powershell-7.4#powershell-execution-policies) allows running a script from the internet.
Request a specific version by including it in the URL:
```pwsh-session
PS> powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/0.8.13/install.ps1 | iex"
```
!!! tip
The installation script may be inspected before use:
=== "macOS and Linux"
```console
$ curl -LsSf https://astral.sh/uv/install.sh | less
```
=== "Windows"
```pwsh-session
PS> powershell -c "irm https://astral.sh/uv/install.ps1 | more"
```
Alternatively, the installer or binaries can be downloaded directly from [GitHub](#github-releases).
</details>
Update uv if already installed - uv releases updates frequently, so make sure you're on the latest.
uv self update
<a id="quickstart-install-and-run-via-pypi-org" href="#quickstart-install-and-run-via-pypi-org">
Quickstart: Install and Run via pypi.org
</a>
Install and run pan-mcp-relay in one line, showing the CLI help:
uvx pan-mcp-relay@latest --help
[!NOTE] While this makes it easy to get started, you can get faster startup times by following the Standard Install
<a id="standard-install" href="#standard-install">
Standard Install
</a>
uv installs to a standard location, which automatically added to your $PATH during uv installation.
On macOS and Linux, this is usually ~/.local/bin.
# Installs to $(uv tool dir)/pan-mcp-relay
uv tool install pan-mcp-relay@latest
# Executable symlinked to $(uv tool dir --bin)/pan-mcp-relay
pan-mcp-relay --help
<a id="install-the-bleeding-edge-development-version" href="#install-the-bleeding-edge-development-version">
Install the bleeding edge development version
</a>
uv tool install https://github.com/PaloAltoNetworks/aisecurity-mcp-relay.git
pan-mcp-relay --help
<a id="upgrade" href="#upgrade">
Upgrade
</a>
If you're using uvx pan-mcp-relay@latest, you're already using the latest version.
Update the installed version with
uv tool install --upgrade pan-mcp-relay
- Install or Upgrade
pan-mcp-relayusinguv:uv tool install --upgrade pan-mcp-relay - Create an
mcp-relay.yamlconfiguration file containing:- Prisma AIRS API Configuration
- Prisma AIRS API Key
- Prisma AIRS AI Profile
- Prisma AIRS API Endpoint (Optional, defaults to United States Prisma AIRS API Endpoint.)
- MCP Servers
- Supports stdio, SSE, and Streamable HTTP MCP Servers.
- Run the MCP Relay Server
- Prisma AIRS API Configuration
<a id="configuration" href="#configuration">
Configuration
</a>
<a id="relay-configuration" href="#relay-configuration">
MCP Relay Server Configuration
</a>
<a id="configuration-precedence" href="#configuration-precedence">
Configuration Precedence
</a>
- CLI Flags
- Environment Variables
- .env file variables
- Configuration File
pan-mcp-relay supports loading a configuration file via --config-file flag, MCP_RELAY_CONFIG_FILE environment variable, or detecting a configuration file on a set of pre-determined locations:
--config-file(-c) CLI flagMCP_RELAY_CONFIG_FILEenvironment variable./mcp-relay.yaml~/.mcp-relay.yaml~/.config/pan-mcp-relay/mcp-relay.yaml./mcp-relay.json~/.mcp-relay.json~/.config/pan-mcp-relay/mcp-relay.json
<a id="configuration-file-format" href="#configuration-file-format">
Configuration File Format
</a>
An example MCP Relay Server configuration is available in the repository in examples/config/mcp-relay.yaml
Copy or create a new mcp-relay.yaml file. The only required section is mcpServers: {}.
# mcpRelay section is optional.
# Specify API Key, AI Profile and API Endpoint via CLI flags or environment variables.
mcpRelay:
# Prisma AIRS API Key (required), environment variables supported
apiKey: |
${PRISMA_AIRS_API_KEY}
# Prisma AIRS AI Profile Name or ID (required), environment variables supported
aiProfile: |
your-ai-profile-name-or-id
# Endpoint Optional, default shown
# endpoint: |
# https://service.api.aisecurity.paloaltonetworks.com
# mcpServers section requires at least one MCP Server.
mcpServers:
# Example / demo MCP Servers
homebrew:
command: brew
args:
- mcp-server
# Optional working directory, supports environment variables
cwd: "${HOMEBREW_PREFIX}"
# optional environment variables, inherited from the parent shell
env:
HOMEBREW_NO_ENV_HINTS: 1
fetch:
command: uvx
args:
- mcp-server-fetch
# filesystem:
# command: npx
# args:
# - -y
# - '@modelcontextprotocol/server-filesystem'
# - /var/tmp/change/this/path
# # Command line arguments support environment variables
# - "${XDG_PICTURES_DIR}"
<a id="environment-variables" href="#environment-variables">
Environment Variables
</a>
In addition to CLI flags and the configuration file, pan-mcp-relay supports setting configuration through the following environment variables:
# Required for Prisma AIRS API
PRISMA_AIRS_API_KEY=YOUR_API_KEY
# Required for Prisma AIRS API
PRISMA_AIRS_AI_PROFILE=YOUR_AI_PROFILE_NAME
# Optional, default is https://service.api.aisecurity.paloaltonetworks.com
PRISMA_AIRS_API_ENDPOINT=https://service.api.aisecurity.paloaltonetworks.com
# See https://pan.dev/prisma-airs/scan/api/#scan-api-endpoints for additional regional API endpoints
# Defaults shown for all other Environment Variables
MCP_RELAY_CONFIG_FILE=~/.config/pan-mcp-relay/mcp-relay.yaml
MCP_RELAY_TRANSPORT=stdio
# Host For SSE Transport Mode
MCP_RELAY_HOST=127.0.0.1
# Port for SSE Transport Mode
MCP_RELAY_TOOL_CACHE_TTL=86400
MCP_RELAY_MAX_SERVERS=32
MCP_RELAY_MAX_TOOLS=256
# Path to optional .env file
MCP_RELAY_DOTENV=
# Supports $PATH-style colon-separated list and environment variables. Directory entries will search for a file named `.env`
# MCP_RELAY_DOTENV=$HOME/.env:~/.config/pan-mcp-relay:$PWD
MCP_RELAY_LOG_LEVEL=INFO
<a id="mcp-client-configuration" href="#mcp-client-configuration">
MCP Client Configuration
</a>
Create a new config file mcp-relay.json to use with your MCP Client (IDE, Chat Client, or Agent).
{
"mcpServers": {
"pan-mcp-relay": {
"command": "uvx",
"args": [
"pan-mcp-relay@latest"
]
}
}
}
For faster startup times, install/upgrade pan-mcp-relay using uv tool install --upgrade pan-mcp-relay, and use the command name directly:
{
"mcpServers": {
"pan-mcp-relay": {
"command": "pan-mcp-relay"
}
}
}
[!CAUTION]
pan-mcp-relayshould be the only MCP Server listed in your MCP Client configuration file.Any other MCP Servers listed here will NOT be visible to the security capabilities of the MCP Relay.
Optionally, specify additional environment variables or CLI flags:
{
"mcpServers": {
"pan-mcp-relay": {
"command": "uvx",
"args": [
"pan-mcp-relay",
"--config-file",
"~/.config/pan-mcp-relay/mcp-relay.yaml"
],
"env": {
"PRISMA_AIRS_AI_PROFILE": "your-ai-profile",
"MCP_RELAY_LOG_LEVEL": "WARNING"
}
}
}
}
<a id="transport-options" href="#transport-options">
Transport Options
</a>
The MCP Relay Server supports the standard MCP transport mechanisms:
stdioTransport: For local process communication (default)StreamableTransport: For HTTP-based communication, supporting SSE (Server-Sent Events).SSETransport: For backwards compatibility with SSE-only MCP Clients.
<a id="usage" href="#usage">
Usage
</a>
<a id="pan-mcp-relay-cli-usage" href="#pan-mcp-relay-cli-usage">
pan-mcp-relay CLI Usage
</a>
pan-mcp-relay --help
Usage: pan-mcp-relay [OPTIONS] COMMAND [ARGS]...
Run the MCP Relay Server.
Options:
-k, --api-key TEXT Prisma AIRS API Key [PRISMA_AIRS_API_KEY=]
-e, --api-endpoint TEXT Prisma AIRS API Endpoint [PRISMA_AIRS_API_ENDPOINT=]
-p, --ai-profile TEXT Prisma AIRS AI Profile Name or ID [PRISMA_AIRS_AI_PROFILE=]
-c, --config-file FILE Path to configuration file (yaml, json) [MCP_RELAY_CONFIG_FILE=]
-t, --transport [stdio|sse|http]
Transport protocol to use [MCP_RELAY_TRANSPORT=] [default: stdio]
-h, --host TEXT Host for HTTP/SSE server [MCP_RELAY_HOST=] [default: 127.0.0.1]
-p, --port INTEGER Port for HTTP/SSE server [MCP_RELAY_PORT=] [default: 8000]
-TTL, --tool-registry-cache-ttl INTEGER
Tool registry cache TTL (in seconds) [MCP_RELAY_TOOL_CACHE_TTL=] [default: 86400]
-MS, --max-mcp-servers INTEGER Maximum number of downstream MCP servers to allow [MCP_RELAY_MAX_SERVERS=] [default: 32]
-MT, --max-mcp-tools INTEGER Maximum number of MCP tools to allow [MCP_RELAY_MAX_TOOLS=] [default: 256]
--env TEXT Use ./.env file, or specify colon separated path to .env file(s)or directories containing .env files. [MCP_RELAY_DOTENV=]
--system-ca Use System CA instead of Mozilla CA Bundle
--capath FILE Path to Custom Trusted CA bundle
-d, --debug Extremely verbose logging output (DEBUG)
-v, --verbose Verbose logging output (INFO, DEFAULT)
-q, --quiet Minimal logging output (WARNING)
--show-config, --dump-config Show configuration and exit [MCP_RELAY_SHOW_CONFIG=]
-V, --version Show the version and exit.
--help Show this message and exit.
Commands:
json-schema Print Configuration File JSON Schema, optionally to a file.
run Run the MCP Relay Server.
show-config Print MCP Relay Configuration and exit.
<a id="running-the-relay-server" href="#running-the-relay-server">
Running the Relay Server
</a>
<a id="stdio-transport-default" href="#stdio-transport-default">
stdio transport (default)
</a>
# Run with stdio transport with automatic configuration detection (default)
uvx pan-mcp-relay
# Specify a configuration file
uvx pan-mcp-relay --config-file ~/.config/pan-mcp-relay/mcp-relay.yaml
# Specify a path to a custom .env file
uvx pan-mcp-relay --env /var/run/secrets/.env
<a id="streamable-http-transport" href="#streamable-http-transport">
Streamable HTTP transport
</a>
# Run with Streamable transport
pan-mcp-relay --transport=http
<a id="sse-transport" href="#sse-transport">
SSE transport
</a>
# Run with SSE transport
pan-mcp-relay --transport=sse
<a id="legal" href="#legal">
Legal
</a>
Copyright © 2025, Palo Alto Networks
Licensed under the Polyform Internal Use License 1.0.0 (the "License"); you may not use this file except in compliance with the License.
You may obtain a copy of the License at:
https://polyformproject.org/licenses/internal-use/1.0.0
(or)
https://github.com/polyformproject/polyform-licenses/blob/76a278c4/PolyForm-Internal-Use-1.0.0.md
As far as the law allows, the software comes as is, without any warranty or condition, and the licensor will not be liable to you for any damages arising out of these terms or the use or nature of the software, under any kind of legal claim.
<!---Protected_by_PANW_Code_Armor_2024 - Y3ByfC9haWZ3L29wZW5zb3VyY2UvYWlzZWN1cml0eS1tY3AtcmVsYXl8MjE5NDV8bWFpbg== --->
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。