Integrations

flow integrates with popular CI/CD platforms, AI assistants, and containerized environments to bring your automation anywhere.

AI Assistant Integration

For a full guide covering MCP, llms.txt, JSON schemas, and Claude Code integration, see AI Tools.

Model Context Protocol (MCP)

Connect flow to AI assistants through the local Model Context Protocol server for natural language workflow management. The flow MCP server enables AI assistants to discover, understand, and execute your flow workflows through conversational interfaces.

Basic Usage

Add the MCP server command to your favorite MCP client:

{
  "mcpServers": {
    "flow": {
      "command": "flow",
      "args": ["mcp"]
    }
  }
}

The server uses stdio transport and provides AI assistants with:

Available Tools:

• get_info - Get flow information, schemas, and current context
• execute - Execute flow workflows
• list_workspaces - List all registered workspaces
• get_workspace - Get details about a specific workspace
• get_workspace_config - Get the full configuration for a specific workspace
• switch_workspace - Change the current workspace
• list_executables - List and filter executables across workspaces
• get_executable - Get detailed information about an executable
• get_execution_logs - Retrieve recent execution logs
• sync_executables - Sync workspace and executable state
• write_flowfile - Create or update a flow file in a workspace

Available Resources:

• flow://workspace/{name} - Workspace metadata and configuration as JSON
• flow://executable/{workspace}/{namespace}/{name} - Executable definition and metadata as JSON
• flow://flowfile/{path} - Raw flowfile YAML content
• flow://logs/{run_id} - Output of a specific execution run as plain text

Available Prompts:

• generate_executable - Generate flow executable configurations
• generate_project_executables - Generate complete project automation sets
• debug_executable - Debug failing executables
• migrate_automation - Convert existing automation to flow
• explain_flow - Explain flow concepts and usage

NOTE

Learn more about MCP: Visit the Model Context Protocol documentation for client setup and integration details.

CI/CD & Deployment

GitHub Actions

Execute flow workflows directly in your GitHub Actions pipelines with the official action.

name: CI
on: [push]

jobs:
  build:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: flowexec/action@v1
        with:
          executable: 'build app'
          params: |
            ENVIRONMENT=production
          secrets: |
            API_KEY=${{ secrets.API_KEY }}

The action installs the flow CLI, registers your repository as a workspace, and runs the specified executable. It supports Linux, macOS, and Windows runners, multi-workspace setups with git-sourced workspaces, vault secret injection, and structured error codes as outputs.

Complete documentation: Visit the Flow Execute Action on GitHub Marketplace.

Docker

Run flow in containerized environments for CI/CD pipelines or isolated execution.

Basic Usage

# Run with default workspace
docker run -it --rm ghcr.io/flowexec/flow

# Execute specific executable
docker run -it --rm ghcr.io/flowexec/flow validate

Environment Variables

• REPO: Repository URL to clone (defaults to flow's repo)
• BRANCH: Git branch to checkout (optional)
• WORKSPACE: Workspace name to use (defaults to "flow")

Workspace from Git

Automatically clone and configure a workspace:

docker run -it --rm \
  -e REPO=https://github.com/your-org/your-workspace \
  -e BRANCH=main \
  -e WORKSPACE=my-workspace \
  ghcr.io/flowexec/flow exec "deploy app"

Local Workspace

Mount your local workspace:

docker run -it --rm \
  -v $(pwd):/workspaces/my-workspace \
  -w /workspaces/my-workspace \
  -e WORKSPACE=my-workspace \
  ghcr.io/flowexec/flow exec "build app"

In CI/CD Pipelines

Any CI/CD platform that supports Docker can run flow. The key is:

1. Use the Docker image: ghcr.io/flowexec/flow
2. Set environment variables: REPO, WORKSPACE, BRANCH as needed
3. Execute your flow commands: flow exec "your-executable"

Note: While this should work, the Docker integration hasn't been extensively tested. If you try flow with other CI/CD platforms, we'd love to hear about your experience!