Chlorpromazine

You can run and interact with the Chlorpromazine MCP Server to ground automated agents in your current project state, verify facts against official sources, and debug complex problems with structured prompts and tools. This MCP server provides a JSON-RPC based interface over HTTP that lets you list prompts, invoke tools, and generate model responses while keeping security and reliability in mind.

How to use

You connect an MCP client to the server to access prompts and tools. Start by ensuring the server endpoint is reachable at the configured host and port. Use the available prompts to steer the agent toward grounded reasoning and fact-checked answers. Invoke tools to perform documentation lookups, perform sober grounding by reading project files, and generate strategic plans when needed. Use the prompts to request sober thinking, fact-checked answers, or a buzzkill style methodical analysis, and use tools to perform live document searches or project state grounding as required.

How to install

Follow these concrete steps to install and run the Chlorpromazine MCP Server locally. Ensure you meet the prerequisites and provide required environment variables as described in the configuration section.

# Prerequisites
node -v
# Should be Node.js 18 or higher

# Optional but recommended: set up a virtual environment if needed for isolation

# 1) Install dependencies
npm install

# 2) Create a configuration file (example)
cat > .env << 'EOF'
SERPAPI_KEY=your_serpapi_key_here
SITE_FILTER=platform.openai.com,docs.anthropic.com
PORT=3000
API_KEY=your_api_key_here
PLANEXE_MCP_URL=http://127.0.0.1:8001
PLANEXE_MCP_API_KEY=your_planexe_api_key_here
EOF

# 3) Start the MCP server
npm start

Additional sections

Configuration, security, and usage notes below summarize the important details you should know to run and maintain the MCP server in production or development environments.

Configuration notes highlight the required and optional environment variables, the MCP HTTP endpoint, and optional integrations with PlanExe for strategic planning workflows.

Security considerations include input sanitization, rate limiting, timeouts on external calls, secrets masking, and safe error handling. For production deployments, place the server behind an API gateway, enable API key authentication, and configure CORS and security headers according to your security posture.

Tools and prompts are designed to help you verify facts, ground agents in the current project state, and debug complex problems by breaking them into manageable steps. Use sober_thinking to read project files and establish a grounded base, use fact_checked_answer to compare responses against official sources, and use buzzkill to structure systematic analysis. When needed, call kill_trip to perform documentation searches and validate information against trusted sources.

Supported MCP connection methods are exposed via an HTTP endpoint and can also be driven through local runtime commands if provided. The primary HTTP MCP endpoint is http://localhost:3000, and the system supports a complete JSON-RPC 2.0 interface with health checks available at /healthz.

Available tools

sober_thinking

Ground the agent in project reality by reading current files like .env, README.md, and CHANGELOG to prevent hallucinations and ensure factual grounding.

fact_checked_answer

Verify answers against official documentation using SerpAPI searches across configured sites to ensure accuracy.

buzzkill

Provide structured analysis to break down complex problems into manageable steps for systematic troubleshooting.

kill_trip

Perform documentation searches with SerpAPI to verify information when the user questions accuracy or requests to check official sources.

strategic_plan

Generate comprehensive strategic plans by invoking the PlanExe MCP Server for business planning and multi-step roadmapping.