mcp-bastion-python 1.0.13

Security middleware for MCP servers protecting LLM agents from prompt injection, resource exhaustion, and PII leakage

MCP-Bastion

Enterprise-Grade Security Middleware for the Model Context Protocol

Releases are published to npm and PyPI via GitHub Actions on tag push.

The Model Context Protocol (MCP) has rapidly become the universally accepted standard for connecting AI agents to enterprise databases and APIs. However, this connectivity introduces a massive new attack surface: unpredictable, non-deterministic agentic behavior.

MCP-Bastion is a lightweight, drop-in security middleware designed to wrap around any existing Python or TypeScript MCP server. Instead of relying on passive logging, human-in-the-loop approvals, or third-party APIs, MCP-Bastion provides an active, 100% local defense layer. It intercepts standard JSON-RPC traffic to stop threats before they cross the enterprise boundary.

Under 5ms proxy overhead. MCP-Bastion provides:

• Prompt Injection Defense: Meta PromptGuard runs locally to block adversarial payloads and jailbreaks.
• PII Redaction: Uses Microsoft Presidio to detect and mask PII before it reaches the LLM context.
• Infinite Loop Protection: Token buckets and cycle detection stop runaway agents from burning API budget.

Secure your MCP server without changing business logic.

---

Core Features

Zero-Click Prompt Injection Prevention

Integrates Meta's PromptGuard model locally to detect and block malicious payloads, jailbreaks, and adversarial tokenization before they reach your external tools.

PII Redaction

Microsoft Presidio scans outbound tool results and masks PII (redaction, substitution, generalization).

Infinite Loop and Denial of Wallet Protection

Implements stateful cycle detection and configurable FinOps token-bucket algorithms to automatically terminate runaway agents and prevent massive API bill overruns.

100% Local Execution (Data Privacy)

All security classification and data redaction happen entirely within the local memory space of your server. Sensitive data never leaves your enterprise network for third-party safety evaluations.

Low Latency

Drop-in middleware, under 5ms overhead.

Framework Integration

Hooks into MCP SDKs (TypeScript, Python) and FastMCP via standard middleware. No business logic changes.

All Features

Feature	Description
Prompt injection	Block jailbreaks via Meta PromptGuard
PII redaction	Mask many entity types via Presidio (e.g. SSN, email, phone, credit card, passport, IBAN, medical license, and more—see Presidio docs)
Rate limiting	Max iterations, timeout, token budget
Audit logging	Log who, what, when, blocked/allowed
Content filter	Block paths/code/URLs, plus allowlist and denylist patterns
Circuit breaker	Disable failing tools after N failures
RBAC	Tool-level permissions by role
Schema validation	Validate tool input types
Replay guard	Block duplicate nonces
Cost tracker	Per-session cost budget
Semantic cache	Cache similar queries

Real-Time Dashboard and Alerts

Run the optional dashboard for a live view of requests, blocked count, PII redacted, cost, top tools, and recent alerts:

mcp-bastion dashboard --port 7000
# or: PYTHONPATH=src python dashboard/app.py

URL	What it returns
http://localhost:7000/	Visual dashboard with charts (requests, blocked by reason, top tools, cost by user, recent alerts)
http://localhost:7000/api/metrics	JSON: requests_total, blocked_total, blocked_pct, pii_redacted_total, cost_total, blocked_by_reason, top_tools, cost_by_user, alerts
http://localhost:7000/api/health	{"status": "ok"}
http://localhost:7000/metrics	Prometheus text format for Grafana/Datadog

Dashboard: total requests, blocked count and %, PII redacted, cost; blocked-by-reason bars; top tools; cost by user; recent alerts.

• Alerts: Slack webhook and cost-threshold alerts. See dashboard/README.md.

Documentation: Use Cases, Attacks, Metrics, Tutorials

Doc	Description
docs/index.md	GitHub Pages-ready docs home
docs/DETAILED_TUTORIAL.md	Step-by-step implementation tutorial for new teams
docs/USE_CASES.md	Real use cases: enterprise gateway, LLM products, internal tools, SaaS, compliance
docs/ATTACK_PREVENTION.md	Examples showing how MCP-Bastion prevents real attacks (injection, PII leak, rate exhaustion, path traversal, RBAC, replay)
docs/METRICS.md	Performance overhead (<5ms) and effectiveness metrics (dashboard, Prometheus, OTEL)
docs/TUTORIALS.md	Tutorials: integrating with FastMCP, TypeScript, GitHub MCP, and open-source MCP servers
docs/GITHUB_PAGES.md	Publish docs as a GitHub Pages website from this same repo

One-Line Docker

docker build -t mcp-bastion/proxy .
docker run -p 8080:8080 mcp-bastion/proxy

MCP endpoint: http://localhost:8080/mcp. Use docker-compose up -d for proxy; add --profile with-dashboard for the dashboard. See DOCKER.md.

Policy-as-Code (bastion.yaml)

Single config file controls all pillars. Copy bastion.yaml.example to bastion.yaml, then:

from mcp_bastion import build_middleware_from_config
middleware = build_middleware_from_config()

See docs/POLICY_AS_CODE.md.

Tip: set hot_reload.enabled: true in bastion.yaml to apply policy changes without restarting your MCP server when using build_middleware_from_config().

CLI for developers

mcp-bastion validate              # validate bastion.yaml
mcp-bastion serve --http 8080     # run MCP server with config
mcp-bastion dashboard --port 7000 # run metrics dashboard

See docs/CLI.md.

OpenTelemetry

Set OTEL_EXPORTER_OTLP_ENDPOINT to export tool-call spans to OTLP. Install optional deps: pip install mcp-bastion-python[otel]. See docs/OTEL.md.

Webhook alerts

Use Slack webhook, a single generic webhook (webhook_url or BASTION_WEBHOOK_URL), or multiple URLs (alerts.webhooks in bastion.yaml). Each blocked event can POST to your endpoint (e.g. PagerDuty, custom API). Webhook delivery now supports retry/backoff controls in bastion.yaml (retry_attempts, retry_backoff_seconds, retry_backoff_max_seconds, timeout_seconds).

---

Why MCP-Bastion (Competitive Comparison)

Early security packages (mcp-guardian, mcp-shield) focus on logging or static scanning. MCP-Bastion adds an active defense layer.

1. Active Defense vs. Passive Logging

The Competition	MCP-Bastion
Tools like mcp-guardian focus on tracing, logging, human-in-the-loop approvals.	Automated interception. MCP-Bastion scrubs PII before it leaves the server.

2. Local Inference vs. Third-Party APIs

The Competition	MCP-Bastion
Many guardrail proxies send prompts to external APIs to check for malice.	PromptGuard-86M and Presidio run locally. Data stays on your network.

3. Stateful Denial of Wallet Protection

The Competition	MCP-Bastion
Most tools focus on static vulns or basic rate limits.	Tracks tool call history per session. Stops runaway loops before they burn API budget.

4. Drop-in Middleware vs. Standalone Gateway

The Competition	MCP-Bastion
Some solutions need standalone proxy servers.	Library hooks into server.setRequestHandler (TS) or middleware (Python). No extra infra.

---

Structure

Path	Description
src/mcp_bastion/	Python package: PromptGuard, Presidio, rate limiting, RBAC, etc.
packages/core/	TypeScript package: rate limiting in-process; prompt/PII via sidecar (MCP_BASTION_URL)
examples/	Python examples (examples/README.md)
dashboard/	Real-time dashboard UI and metrics API (dashboard/README.md)
bastion.yaml.example	Policy-as-code sample; copy to bastion.yaml (docs/POLICY_AS_CODE.md)
scripts/validate_checklist.py	Enterprise validation runner
VALIDATION_CHECKLIST.md	Validation guide and MCP Inspector steps
SETUP_GUIDE.md	Setup, config, and validation
DOCKER.md	Docker one-line run and compose

Example Files

File	Purpose
examples/python_server_example.py	Minimal middleware chain
examples/full_demo.py	All 11 features (rate limit, PII, RBAC, etc.)
examples/llm_server.py	Shared MCP server for LLM clients
examples/llm_openai_example.py	OpenAI
examples/llm_claude_example.py	Claude
examples/llm_gemini_example.py	Gemini
examples/llm_mistral_example.py	Mistral
examples/llm_grok_example.py	Grok (xAI)
examples/server_with_config.py	Policy-as-code (bastion.yaml)

Installation

Python

uv add mcp-bastion-python
# or
pip install mcp-bastion-python

Prerequisites (recommended)

• PII redaction: Presidio expects the spaCy English model. After install, run:
  python -m spacy download en_core_web_sm
  Without it, PII analysis can fail at runtime.
• Policy-as-Code (bastion.yaml): install YAML support:
  pip install mcp-bastion-python[policy]
  (adds pyyaml; otherwise you may get ImportError when loading policy files).
• PromptGuard fail-open: if the PromptGuard model fails to load or inference errors, MCP-Bastion allows the request and logs a warning. Treat this as a security degradation and fix the model/runtime before production.

The PyPI wheel ships the full mcp_bastion tree (including config, cli, otel, dashboard metrics, and alert sinks). If you use an older wheel that omits modules, upgrade to the current release.

TypeScript

npm install @mcp-bastion/core

npm

Framework Integrations

Drop-in security for your favorite LLM framework. Each package auto-installs mcp-bastion-python:

pip install mcp-bastion-langchain      # LangChain agents and tools
pip install mcp-bastion-openai         # OpenAI GPT API calls
pip install mcp-bastion-anthropic      # Anthropic Claude API calls
pip install mcp-bastion-bedrock        # AWS Bedrock runtime
pip install mcp-bastion-gemini         # Google Gemini
pip install mcp-bastion-crewai         # CrewAI agent crews
pip install mcp-bastion-llamaindex     # LlamaIndex RAG pipelines
pip install mcp-bastion-groq           # Groq inference
pip install mcp-bastion-mistral        # Mistral AI
pip install mcp-bastion-cohere         # Cohere
pip install mcp-bastion-azure          # Azure OpenAI Service
pip install mcp-bastion-vertexai       # Google Cloud Vertex AI
pip install mcp-bastion-huggingface    # Hugging Face Inference
pip install mcp-bastion-deepseek       # DeepSeek AI
pip install mcp-bastion-together       # Together AI
pip install mcp-bastion-fireworks      # Fireworks AI
pip install mcp-bastion-fastmcp       # FastMCP servers

Package	Protects	Version	Downloads
mcp-bastion-langchain	LangChain	0.1.0	stats
mcp-bastion-openai	OpenAI GPT	0.1.0	stats
mcp-bastion-anthropic	Anthropic Claude	0.1.0	stats
mcp-bastion-bedrock	AWS Bedrock	0.1.0	stats
mcp-bastion-gemini	Google Gemini	0.1.0	stats
mcp-bastion-crewai	CrewAI	0.1.0	stats
mcp-bastion-llamaindex	LlamaIndex	0.1.0	stats
mcp-bastion-groq	Groq	0.1.0	stats
mcp-bastion-mistral	Mistral AI	0.1.0	stats
mcp-bastion-cohere	Cohere	0.1.0	stats
mcp-bastion-azure	Azure OpenAI	0.1.1	stats
mcp-bastion-vertexai	Vertex AI	0.1.0	stats
mcp-bastion-huggingface	Hugging Face	0.1.1	stats
mcp-bastion-deepseek	DeepSeek AI	0.1.1	stats
mcp-bastion-together	Together AI	0.1.1	stats
mcp-bastion-fireworks	Fireworks AI	0.1.1	stats
mcp-bastion-fastmcp	FastMCP servers	0.1.0	stats

Publish (PyPI / npm)

• PyPI: python -m build && twine upload dist/* (or use GitHub Actions on tag).
• npm: From repo root, cd packages/core && npm publish --access public (or use Trusted Publishers).
• Version is set in pyproject.toml (Python), packages/core/package.json (npm), and server.json (MCP registry). Bump before releasing.

Developer Guide

Integration examples for Python and TypeScript.

---

Quick Start (Python)

Add MCP-Bastion to an existing MCP server in three steps:

from mcp_bastion import MCPBastionMiddleware, compose_middleware

# 1. Create the security middleware
bastion = MCPBastionMiddleware(
    enable_prompt_guard=True,
    enable_pii_redaction=True,
    enable_rate_limit=True,
)

# 2. Compose with your middleware chain (Bastion runs first)
middleware = compose_middleware(bastion)

# 3. Pass the composed middleware to your MCP server
# (integration depends on your server framework)

Examples:

Example	Description
examples/python_server_example.py	Basic middleware chain
examples/full_demo.py	All features: add, PII, rate limit, prompt injection
examples/llm_openai_example.py	MCP server for OpenAI
examples/llm_claude_example.py	MCP server for Claude
examples/llm_gemini_example.py	MCP server for Gemini
examples/llm_mistral_example.py	MCP server for Mistral
examples/llm_grok_example.py	MCP server for Grok (xAI, HTTP only)

# Windows: $env:PYTHONPATH="src"; python examples/full_demo.py
# Linux/Mac: PYTHONPATH=src python examples/full_demo.py

LLM integration: See docs/LLM_INTEGRATION.md for copy-paste config for OpenAI, Claude, Gemini, Mistral, and Grok.

Enterprise validation:

PYTHONPATH=src python scripts/validate_checklist.py

See VALIDATION_CHECKLIST.md and SETUP_GUIDE.md.

---

Python Tutorial: FastMCP Server

FastMCP server with MCP-Bastion.

Step 1: Install dependencies

pip install mcp mcp-bastion-python

Step 2: Create your server file (server.py)

from mcp.server.fastmcp import FastMCP
from mcp_bastion import MCPBastionMiddleware, compose_middleware

# Create the MCP server
mcp = FastMCP("My Secure Server")

# Create MCP-Bastion middleware
# It intercepts tool calls and resource reads before they execute
bastion = MCPBastionMiddleware(
    enable_prompt_guard=True,   # Block malicious prompts via PromptGuard
    enable_pii_redaction=True,  # Mask PII in outgoing content
    enable_rate_limit=True,     # Cap at 15 iterations, 60s timeout
)

# Compose middleware chain (pass to your server's middleware config if supported)
middleware = compose_middleware(bastion)

# Register a tool (protected when middleware is wired into your server)
@mcp.tool()
def get_weather(city: str) -> str:
    """Get weather for a city."""
    return f"Weather in {city}: 22C, sunny"

# Resource (PII redacted)
@mcp.resource("user://profile/{user_id}")
def get_profile(user_id: str) -> str:
    """Get user profile. PII redacted."""
    return f"User {user_id}: John Doe, SSN 123-45-6789, [e-mail]"

if __name__ == "__main__":
    mcp.run(transport="streamable-http")

Step 3: Run the server

python server.py

MCP-Bastion:

• Scans tool args for prompt injection
• Redacts PII from resource responses
• Blocks sessions over 15 calls or 60s

Alternative: Policy-as-Code

Use bastion.yaml instead of code. Copy bastion.yaml.example to bastion.yaml, then:

from mcp_bastion import build_middleware_from_config
middleware = build_middleware_from_config()

See docs/POLICY_AS_CODE.md and examples/server_with_config.py.

---

Python: Custom Rate Limits

Custom config example:

from mcp_bastion import MCPBastionMiddleware
from mcp_bastion.pillars.rate_limit import TokenBucketRateLimiter
from mcp_bastion.pillars.prompt_guard import PromptGuardEngine

# Stricter limits
rate_limiter = TokenBucketRateLimiter(
    max_iterations=10,
    timeout_seconds=30,
    token_budget=25_000,
)

# Higher threshold = fewer blocks, more risk
prompt_guard = PromptGuardEngine(threshold=0.92)

bastion = MCPBastionMiddleware(
    prompt_guard=prompt_guard,
    rate_limiter=rate_limiter,
    enable_prompt_guard=True,
    enable_pii_redaction=True,
    enable_rate_limit=True,
)

# Disable PII redaction if your data has no PII
bastion_no_pii = MCPBastionMiddleware(enable_pii_redaction=False)

---

Python: Custom Middleware

Extend Middleware to add logging, metrics, or custom logic:

from mcp_bastion.base import Middleware, MiddlewareContext, compose_middleware

class LoggingMiddleware(Middleware):
    async def on_message(self, context, call_next):
        result = await call_next(context)
        # log method, elapsed, etc.
        return result

middleware = compose_middleware(bastion, LoggingMiddleware())

See examples/full_demo.py for a complete example.

---

TypeScript: Wrap an MCP Server

Step 1: Install dependencies

npm install @modelcontextprotocol/sdk @mcp-bastion/core

Step 2: Create your server (server.ts)

import { Server } from "@modelcontextprotocol/sdk/server/index.js";
import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";
import {
  wrapWithMcpBastion,
  wrapCallToolHandler,
} from "@mcp-bastion/core";

const server = new Server({ name: "my-mcp-server", version: "1.0.0" });

// Wrap the server with MCP-Bastion (rate limiting only by default)
// For prompt injection and PII, run the Python sidecar and set sidecarUrl
wrapWithMcpBastion(server, {
  enableRateLimit: true,
  maxIterations: 15,
  timeoutMs: 60_000,
  // Optional: enable ML features via Python sidecar
  sidecarUrl: process.env.MCP_BASTION_SIDECAR || "",
  enablePromptGuard: !!process.env.MCP_BASTION_SIDECAR,
  enablePiiRedaction: !!process.env.MCP_BASTION_SIDECAR,
});

// Register tools (handlers are automatically wrapped)
server.setRequestHandler("tools/call" as any, async (request) => {
  if (request.params?.name === "get_weather") {
    return {
      content: [{ type: "text", text: "Sunny, 22C" }],
      isError: false,
    };
  }
  throw new Error("Unknown tool");
});

async function main() {
  const transport = new StdioServerTransport();
  await server.connect(transport);
}

main();

Step 3: Run with rate limiting only

npx tsx server.ts

Step 4: Run with full ML features (Python sidecar)

For prompt injection and PII redaction, run a Python HTTP service that exposes /prompt-guard and /pii-redact endpoints (see the Python package for sidecar implementation). Then:

# Start the Python sidecar, then the TypeScript server
MCP_BASTION_SIDECAR=http://localhost:8000 npx tsx server.ts

---

TypeScript: Wrap Individual Handlers

Wrap specific handlers only:

import {
  wrapCallToolHandler,
  wrapReadResourceHandler,
} from "@mcp-bastion/core";
import {
  CallToolRequestSchema,
  ReadResourceRequestSchema,
} from "@modelcontextprotocol/sdk/types.js";

// Wrap only the tool handler
const safeToolHandler = wrapCallToolHandler(
  async (request) => {
    // Your tool logic
    return { content: [{ type: "text", text: "OK" }], isError: false };
  },
  { enableRateLimit: true, maxIterations: 10 }
);

// Wrap only the resource handler (for PII redaction)
const safeResourceHandler = wrapReadResourceHandler(
  async (request) => {
    const contents = await fetchResource(request.params.uri);
    return { contents };
  },
  { sidecarUrl: "http://localhost:8000", enablePiiRedaction: true }
);

server.setRequestHandler(CallToolRequestSchema, safeToolHandler);
server.setRequestHandler(ReadResourceRequestSchema, safeResourceHandler);

---

Configuration Reference

Option	Python	TypeScript	Default	Description
enable_prompt_guard	Yes	Yes	True (Python) / False (TS)	Block malicious prompts via PromptGuard
enable_pii_redaction	Yes	Yes	True (Python) / False (TS)	Mask PII in outgoing content
enable_rate_limit	Yes	Yes	True	Enforce iteration and timeout caps
max_iterations	Via TokenBucketRateLimiter	Yes	15	Max tool calls per session
timeout_seconds / timeoutMs	Via TokenBucketRateLimiter	Yes	60	Session timeout
token_budget	Via TokenBucketRateLimiter	-	50,000	FinOps token cap per request
sidecarUrl	-	Yes	""	Python sidecar URL for ML features
threshold	Via PromptGuardEngine	-	0.85	Malicious probability cutoff
setLogLevel	-	Yes	"info"	TypeScript: "debug" | "info" | "warn" | "error"

---

Error Handling

When MCP-Bastion blocks a request, it returns standard MCP/JSON-RPC errors:

Code	Exception	When
-32001	PromptInjectionError	Tool args contain jailbreak/injection
-32002	RateLimitExceededError	Session exceeds iteration or timeout limit
-32003	TokenBudgetExceededError	Session exceeds token budget
-32004	CircuitBreakerOpenError	Tool’s circuit breaker is open after failures
-32005	ContentFilterError	Content filter matched a blocked pattern
-32006	RBACError	Caller not allowed to use this tool
-32007	SchemaValidationError	Tool arguments failed schema validation
-32008	ReplayAttackError	Duplicate nonce / replay detected
-32009	CostBudgetExceededError	Session cost budget exceeded

# Python: exceptions
from mcp_bastion.errors import (
    PromptInjectionError,
    RateLimitExceededError,
    TokenBudgetExceededError,
    CircuitBreakerOpenError,
    ContentFilterError,
    RBACError,
    SchemaValidationError,
    ReplayAttackError,
    CostBudgetExceededError,
)
import logging
logger = logging.getLogger(__name__)

try:
    result = await middleware(context, call_next)
except (
    PromptInjectionError,
    RateLimitExceededError,
    TokenBudgetExceededError,
    CircuitBreakerOpenError,
    ContentFilterError,
    RBACError,
    SchemaValidationError,
    ReplayAttackError,
    CostBudgetExceededError,
) as e:
    logger.warning("blocked: %s", e.to_mcp_error())

// TypeScript: handlers return isError: true
import { logger, setLogLevel } from "@mcp-bastion/core";
setLogLevel("debug");  // optional: "debug" | "info" | "warn" | "error"
const result = await guardedHandler(request);
if (result.isError) {
  logger.error("blocked", result.content);
}

---

Testing

MCP Inspector:

# Start your guarded server
python server.py   # or: npx tsx server.ts

# In another terminal, launch the Inspector
npx -y @modelcontextprotocol/inspector

Connect via HTTP (http://localhost:8000/mcp) or stdio, then:

1. List tools and call one with benign arguments (should succeed)
2. Call a tool with "Ignore previous instructions" (should be blocked)
3. Trigger 16+ tool calls in one session (should hit rate limit)

---

Testing

# Python (PYTHONPATH=src on Windows: $env:PYTHONPATH="src")
pytest tests/ -v

# TypeScript
npm run test --workspace=@mcp-bastion/core

# Full validation checklist (build, pillars, latency)
PYTHONPATH=src python scripts/validate_checklist.py

# MCP Inspector (manual)
npx -y @modelcontextprotocol/inspector

Third-Party Components

See NOTICE for licenses. MCP-Bastion uses Meta Llama Prompt Guard 2 (Llama 4 Community License) and Microsoft Presidio. For OWASP-relevant mitigations, dependency audit, and reporting vulnerabilities, see docs/SECURITY.md.

License

MCP-Bastion is distributed under the MCP-Bastion Community and Commercial License.

• Non-commercial use is permitted with required attribution/citation.
• Commercial use requires a separate paid agreement.

See:

• LICENSE
• COMMERCIAL_LICENSE.md
• CITATION.cff