bowtie-mcp
MCP server for managing Bowtie clusters, enabling natural language control of policies, devices, DNS, users, and more.
README
Bowtie MCP Server
An MCP server for the Bowtie control plane. Manage your Bowtie cluster — policies, devices, DNS, users, and more — using natural language through your LLM of choice or any MCP-capable client.
Install
pip install bowtie-mcp
Or run without installing using uvx:
uvx bowtie-mcp
Quick Start
Claude Code
claude mcp add bowtie \
-e BOWTIE_HOST="https://bt0.example.com" \
-e BOWTIE_USERNAME="admin@example.com" \
-e BOWTIE_PASSWORD="your-password" \
-- uvx bowtie-mcp
Claude Desktop
Add to your claude_desktop_config.json:
{
"mcpServers": {
"bowtie": {
"command": "uvx",
"args": ["bowtie-mcp"],
"env": {
"BOWTIE_HOST": "https://bt0.example.com",
"BOWTIE_USERNAME": "admin@example.com",
"BOWTIE_PASSWORD": "your-password"
}
}
}
}
Cursor / Windsurf
Add to .cursor/mcp.json or your editor's MCP config:
{
"mcpServers": {
"bowtie": {
"command": "uvx",
"args": ["bowtie-mcp"],
"env": {
"BOWTIE_HOST": "https://bt0.example.com",
"BOWTIE_USERNAME": "admin@example.com",
"BOWTIE_PASSWORD": "your-password"
}
}
}
}
HTTP Streaming Mode
For clients that connect over HTTP instead of spawning a local stdio process, start the server in one of FastMCP's HTTP transports:
uvx bowtie-mcp --transport streamable-http --port 8080
This binds to 127.0.0.1:8080 and exposes the recommended Streamable HTTP endpoint at http://127.0.0.1:8080/mcp.
Legacy SSE is also available:
uvx bowtie-mcp --transport sse --port 8080
The SSE endpoint is http://127.0.0.1:8080/sse. During the handshake, FastMCP tells the client to post subsequent messages to the matching http://127.0.0.1:8080/messages/ endpoint.
Authentication
The server authenticates to your Bowtie controller using admin credentials — the same approach used by the Bowtie Terraform provider.
| Variable | Required | Description |
|---|---|---|
BOWTIE_HOST |
Yes | Controller HTTPS endpoint (e.g., https://bt0.example.com) |
BOWTIE_USERNAME |
Yes | Admin email address |
BOWTIE_PASSWORD |
Yes | Admin password |
BOWTIE_MCP_SKIP_CONFIRMATION |
No | Set to true to skip write confirmation prompts |
Credentials are used by the local MCP server process to authenticate with your controller. They are never sent to or visible by the LLM.
What You Can Do
Ask your LLM things like:
- "List all pending devices and approve the ones from the engineering team"
- "Create a policy that gives the sales team access to the CRM on 10.0.5.0/24 port 443"
- "Why can't device X reach the production database?"
- "Audit our policies for anything overly permissive"
- "Set up DNS for internal.example.com pointing at 10.0.0.53"
Available Tools
Read (30+ tools) — List and inspect users, devices, device groups, user groups, policies, resources, resource groups, DNS configs, sites, collections, block lists, threat categories, controllers, route exclusions, and API keys.
Write (25+ tools) — Create, update, and delete all of the above. All write operations require explicit confirmation before executing (see below).
Guided Workflows (4 prompts):
| Prompt | Description |
|---|---|
onboard-team |
Create a user group, resources, resource group, and access policy |
audit-policies |
Review all policies for overly permissive rules |
troubleshoot-access |
Diagnose why a user/device can or can't reach a destination |
review-pending-devices |
Batch approve or reject pending devices |
Write Confirmation
All write tools use a two-step confirmation pattern. The first call returns a preview of the planned change. Call again with confirm=true to execute. This prevents accidental mutations.
Tool responses are returned as structured JSON-compatible objects rather than pre-serialized JSON strings, so MCP clients can inspect fields directly.
Set BOWTIE_MCP_SKIP_CONFIRMATION=true to bypass this for automation or CI workflows.
Development
git clone https://github.com/bowtieworks/bowtie-mcp.git
cd bowtie-mcp
pip install -e ".[dev]"
pytest
See CONTRIBUTING.md for more details.
License
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.