MCP Server¶

Tenets is a free, open source, 100% local MCP server that solves two critical problems for AI coding:

1. Intelligent Code Context — NLP-powered ranking finds and aggregates the most relevant code automatically
2. Automatic Guiding Principles — Your tenets (coding standards, rules) are injected into every prompt, preventing context drift

Integrates natively with Cursor, Claude Desktop, Windsurf, and custom AI agents.

Learn more: Model Context Protocol Explained · Why Context Matters

Installation¶

pip install tenets[mcp]

Quick Start¶

Start the Server¶

# Default: stdio transport for local IDE integration
tenets-mcp

# SSE transport for web clients
tenets-mcp --transport sse --port 8080

# HTTP transport for remote deployment
tenets-mcp --transport http --port 8080

Verify Installation¶

tenets-mcp --version

Python 3.14 compatibility¶

• Use tenets >= 0.7.0 on Python 3.14 to avoid the prior import recursion in tenets.core.
• No extra flags needed; stdio/SSE/HTTP transports all work on 3.14+.

IDE Configuration¶

Claude Desktop¶

Add to your Claude Desktop configuration (~/Library/Application Support/Claude/claude_desktop_config.json on macOS):

{
  "mcpServers": {
    "tenets": {
      "command": "tenets-mcp",
      "args": []
    }
  }
}

If you installed tenets in a virtual environment, use the full path:

{
  "mcpServers": {
    "tenets": {
      "command": "/path/to/venv/bin/tenets-mcp",
      "args": []
    }
  }
}

Cursor¶

Add to your Cursor MCP settings (Settings → MCP Servers):

{
  "tenets": {
    "command": "tenets-mcp",
    "args": []
  }
}

Windsurf / Codeium¶

Configure the MCP server in Windsurf settings:

{
  "mcp": {
    "servers": {
      "tenets": {
        "command": "tenets-mcp"
      }
    }
  }
}

VS Code with MCP Extension¶

If using an MCP extension for VS Code:

{
  "mcp.servers": {
    "tenets": {
      "command": "tenets-mcp",
      "transport": "stdio"
    }
  }
}

Available Tools¶

The MCP server exposes these tools to AI assistants:

Context Tools¶

Analysis Tools¶

Session Tools¶

Tenet Tools¶

Available Resources¶

Resources are read-only data the AI can access:

Available Prompts¶

Prompts are reusable templates for common tasks:

Tool-to-Prompt Matrix¶

Common development tasks and which tools to use:

Example Usage¶

Once configured, ask your AI assistant:

"Use tenets to find relevant files for implementing user authentication"

The AI will call the distill tool and return ranked, optimized context.

"Create a session called 'auth-feature' and pin the auth folder"

The AI will use session_create and session_pin_folder.

"Add a tenet: Always validate user input before processing"

The AI will use tenet_add to create a guiding principle.

Tool Responses (Examples)¶

distill¶

{
  "context": "# File: src/auth.py\n...",
  "token_count": 45000,
  "files": ["src/auth.py", "src/user.py"],
  "files_summarized": ["src/utils.py"],
  "metadata": {"mode": "balanced", "total_scanned": 150}
}

rank_files¶

{
  "files": [
    {"path": "src/auth.py", "score": 0.85},
    {"path": "src/user.py", "score": 0.72}
  ],
  "total_scanned": 150,
  "mode": "balanced"
}

session_create¶

{
  "id": "sess_abc123",
  "name": "auth-feature",
  "created_at": "2025-12-04T10:00:00"
}

Error Semantics¶

Tenets returns clear errors that AI agents can act on:

• Path errors (missing or inaccessible):
• Distill/Rank will raise an error or return an empty result — agents should retry with a valid path.
• Invalid parameters (mode/format):
• MCP validation prevents tool invocation; agents should correct parameters based on schema.
• Session operations:
• Pinning a nonexistent file returns a structured failure response; agents should prompt to confirm the path.

When in doubt, agents should: 1) Validate inputs against the tool schema, and 2) Ask the user to clarify scope (path, include/exclude patterns, session name).

Remote Deployment¶

For team or cloud deployment, use HTTP transport:

# Start server on all interfaces
tenets-mcp --transport http --host 0.0.0.0 --port 8080

Configure clients to connect:

{
  "mcpServers": {
    "tenets": {
      "url": "http://your-server:8080/mcp"
    }
  }
}

Programmatic Usage¶

from tenets.mcp import create_server

# Create and run server
server = create_server()
server.run(transport="stdio")

# Or with custom config
from tenets.config import TenetsConfig
config = TenetsConfig(max_tokens=150000)
server = create_server(config=config)
server.run(transport="http", port=8080)

Configuration¶

MCP settings can be added to .tenets.yml:

mcp:
  enabled: true

  # Tool availability
  tools:
    distill: true
    rank_files: true
    examine: true
    chronicle: true
    momentum: true
    session: true
    tenet: true

Troubleshooting¶

Server not starting¶

1. Verify MCP dependencies: pip install tenets[mcp]
2. Check Python version: requires 3.9+
3. Try verbose mode: tenets-mcp --verbose

IDE not connecting¶

1. Restart the IDE after configuration changes
2. Check the server is running: tenets-mcp --version
3. Verify the command path is correct in IDE settings
4. Check IDE logs for MCP connection errors

Tools not appearing¶

1. Ensure MCP server is running
2. Refresh MCP connection in IDE
3. Check server logs for initialization errors

MCP Discovery & Marketplaces¶

How AI Assistants Find MCP Servers¶

MCP servers are discovered through configuration files. Each AI assistant has its own configuration format:

Publishing to MCP Directories¶

To make tenets discoverable by other developers:

1. Official MCP Server Registry¶

Submit to the MCP Servers Directory:

# Fork and clone the servers repository
git clone https://github.com/modelcontextprotocol/servers.git

# Add your server to the README
# Submit a pull request

2. Awesome MCP Lists¶

Submit to community-curated lists: - awesome-mcp - Community curated - mcp-directory - Searchable directory

3. Package Managers¶

Tenets is available via pip, making it easy to install:

pip install tenets[mcp]

Local Development Setup¶

For developers working on tenets locally:

# Clone the repository
git clone https://github.com/jddunn/tenets.git
cd tenets

# Install in development mode with MCP support
pip install -e ".[mcp,dev]"

# Run the MCP server
tenets-mcp

Registering with IDEs Automatically¶

Some IDEs support automatic MCP server discovery through package metadata.

pyproject.toml Entry Point¶

Tenets registers an entry point that MCP-aware tools can discover:

[project.scripts]
tenets-mcp = "tenets.mcp.server:main"

Manual Registration¶

If automatic discovery isn't available, add to your IDE config:

Claude Desktop (macOS):

# Open config
open ~/Library/Application\ Support/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "tenets": {
      "command": "tenets-mcp"
    }
  }
}

Cursor: 1. Open Settings (⌘ + ,) 2. Navigate to MCP Servers 3. Add new server with command tenets-mcp

Verifying Installation¶

After configuration, verify the server is recognized:

1. Restart your IDE after config changes
2. Check server status in IDE's MCP panel
3. Test a tool call: Ask "Use tenets to rank files for authentication"

Distributing to Teams¶

For team deployment, create a shared configuration:

{
  "mcpServers": {
    "tenets": {
      "command": "tenets-mcp",
      "env": {
        "TENETS_MAX_TOKENS": "150000",
        "TENETS_RANKING_ALGORITHM": "balanced"
      }
    }
  }
}

Share this configuration file with your team for consistent setup.

Enterprise Features¶

The following features are available with enterprise support:

• Privacy Redaction: Automatically strip API keys, credentials, and PII
• Anonymization: Remove author information for compliance
• Custom ML Models: Fine-tuned models for your codebase
• SSO/SAML: Enterprise authentication
• Audit Logging: Compliance tracking
• Air-gapped Deployment: Offline operation support

Contact us for enterprise inquiries.

Related Resources¶

• Cursor MCP Setup Guide — Detailed Cursor configuration
• Claude Desktop MCP Guide — Configure Claude Desktop
• Quick Start — Get started in 30 seconds
• CLI Reference — Command-line interface documentation
• FAQ — Frequently asked questions
• Model Context Protocol Explained — Technical deep dive into MCP
• Why Context Matters — The importance of intelligent context