CodePipeline

You can manage AWS CodePipeline through this MCP server, enabling you to list pipelines, view states and definitions, trigger executions, handle approvals, and integrate with Windsurf and Cascade for natural language control.

How to use

You use this MCP server with an MCP client like Windsurf to translate natural language requests into CodePipeline actions. After starting the server, you can ask to list pipelines, inspect pipeline states and details, trigger runs, approve or reject manual actions, retry failed stages, view logs, tag resources, and create webhooks. The interface is designed to be consistent with other MCP servers so you can compose queries in plain language and have Cascade perform the corresponding CodePipeline operations.

How to install

Prerequisites: you need Node.js version 14 or later and an AWS account with CodePipeline access. Ensure you have AWS credentials with permissions for CodePipeline, CloudWatch, and IAM for tagging.

# 1) Install Node.js (14+). Verify with:
node -v
npm -v

# 2) Clone the MCP CodePipeline server repository
# (follow the repository's cloning instructions to obtain the source code)
# Example:
# git clone https://github.com/your-org/mcp-codepipeline-server.git
# cd mcp-codepipeline-server

# 3) Install dependencies
npm install

# 4) Create a local environment file
cp .env.example .env

# 5) Configure environment variables in .env
# Example values; replace with your credentials
AWS_REGION=us-east-1
AWS_ACCESS_KEY_ID=your_access_key_id
AWS_SECRET_ACCESS_KEY=your_secret_access_key
PORT=3000

# 6) Build the project (if required by the project setup)
npm run build

# 7) Start the server locally
npm start

# For development with auto-restart (if supported)
npm run dev

Colorful configuration for Windsurf integration

To connect Windsurf to this MCP server, add a Windsurf configuration that points to the MCP runner and includes your AWS credentials. Place the configuration under your Windsurf MCP config. The example below uses the npx approach to run the MCP server from its built distribution.

{
  "mcpServers": {
    "codepipeline": {
      "command": "npx",
      "args": [
        "-y",
        "path/to/mcp-codepipeline-server/dist/index.js"
      ],
      "env": {
        "AWS_REGION": "us-east-1",
        "AWS_ACCESS_KEY_ID": "your_access_key_id",
        "AWS_SECRET_ACCESS_KEY": "your_secret_access_key"
      }
    }
  }
}

Start and run options

You have two convenient ways to run the MCP server, depending on your setup.

Option A: Run with npm (local development). Start the server directly from your project using npm.

npm start

Option B: Run via npx from a built distribution. This assumes you have the built assets available at the specified path.

npx -y path/to/mcp-codepipeline-server/dist/index.js

Configuration notes

For security, do not commit your credentials in version control. Use a dedicated environment file and restrict access to it. When running in production, prefer roles and more restricted credentials, and rotate keys periodically.

Troubleshooting and tips

If you encounter connection or credential errors, verify that your AWS credentials are correct and that the associated IAM user has the necessary permissions for CodePipeline, CloudWatch, and IAM tagging. Ensure the server is reachable on the configured port and that there are no firewall rules blocking access.

Example usage patterns with Windsurf and Cascade

Once connected, you can issue natural language requests such as listing pipelines, checking a pipeline's state, triggering executions, approving actions, or retrieving metrics. Cascade translates these requests into the appropriate MCP tool calls behind the scenes.

Examples

{
  "pipelineName": "my-pipeline",
  "action": "list_pipelines"
}

Available tools

list_pipelines

List all CodePipeline pipelines

get_pipeline_state

Get the current state of a specific pipeline

list_pipeline_executions

List executions for a specific pipeline

trigger_pipeline

Trigger a pipeline execution

stop_pipeline_execution

Stop a pipeline execution

get_pipeline_details

Get the full definition of a pipeline

get_pipeline_execution_logs

Get logs for a pipeline execution

get_pipeline_metrics

Get performance metrics for a pipeline

approve_action

Approve or reject a manual approval action

retry_stage

Retry a failed stage

tag_pipeline_resource

Add or update tags for a pipeline resource

create_pipeline_webhook

Create a webhook for a pipeline