Devpipe

You can run and manage Devpipe pipelines from an MCP server so your AI assistants can list, validate, and execute tasks in your local development workflows. This server lets you interact with devpipe through natural prompts, giving you hands-on control over configurations, runs, and results.

How to use

You interact with the MCP server using an MCP client. Tell it which project you want to operate on, then ask it to list tasks, validate configurations, run pipelines, or analyze results. The server translates your natural language requests into concrete devpipe actions and returns structured results you can inspect.

How to install

Prerequisites you need before you start:

• Node.js 18 or higher

• Access to npm and npx in your shell

Install the MCP server globally using npm

npm install -g devpipe-mcp

Install devpipe-mcp and verify it is available in your PATH

## Configuration and startup

Configure your MCP client to connect to the Devpipe MCP server via a stdio channel. You can provide the following configuration in your MCP setup to run the server locally with npx and the latest package:

{ "mcpServers": { "devpipe": { "command": "npx", "args": [ "-y", "devpipe-mcp@latest" ] } } }

## Using the MCP with your AI assistant

After configuration, your AI assistant can perform common tasks by prompting with natural language. Typical actions include listing tasks, running pipelines, validating configurations, reading run logs, and generating CI configs. Start by specifying the project you want to work with and then request the action you need, such as analyzing a project or generating a config for CI.

## Examples of common prompts

List tasks in a project, run a subset of tasks, or validate a config. You can also request insights like suggested optimizations or security findings from SARIF reports.

## Security and metrics considerations

Review results and SARIF findings from a run, access JUnit/SARIF metrics, and inspect logs for troubleshooting. The MCP server supports commands to parse metrics and inspect dashboards or summary data.

## Shutdown and cleanup

When you are finished, gracefully stop your MCP client and ensure any background tasks or caches are closed. If you are running in a persistent environment, you can keep the MCP server available for subsequent prompts.

## Related tips

Keep your devpipe and MCP client versions in sync. Use the latest compatible versions to ensure all features and prompts work reliably.

## Troubleshooting

If you encounter issues starting or connecting, verify Node.js and npm/npx are accessible, and confirm the npx command resolves correctly to devpipe-mcp@latest.

## Available tools

### list\_tasks

Parse and list all tasks from a config file.

### run\_pipeline

Execute devpipe with specified flags to run a pipeline.

### validate\_config

Validate devpipe configuration files for correctness.

### get\_last\_run

Retrieve results from the most recent pipeline run.

### view\_run\_logs

Read logs from a specific task or the entire pipeline.

### parse\_metrics

Parse JUnit or SARIF metrics files to expose results.

### get\_dashboard\_data

Extract aggregated data from summary data.

### check\_devpipe

Check if devpipe is installed and obtain version information.

### list\_tasks\_verbose

List tasks with verbose execution statistics.

### analyze\_project

Analyze a project to detect technologies and suggest missing tasks.

### generate\_task

Generate a task configuration from a template for a given technology.

### create\_config

Create a complete config.toml from scratch with auto-detected tasks.

### get\_pipeline\_health

Calculate overall pipeline health with trend analysis and recommendations.

### compare\_runs

Compare two pipeline runs to identify changes in failures and performance.

### predict\_impact

Predict which tasks are likely to fail based on changes and history.

### generate\_ci\_config

Generate CI/CD configuration from devpipe config for GitHub Actions or GitLab CI.