Control AI Agent Workflows

Define multi-step workflows for AI agents. MCP Moira guarantees each step executes in order, validates the output structure, and preserves context between steps. Works with Claude Code, Cursor, ChatGPT, and any MCP client.

Get Started

View on GitHub

MCP Protocol Native

Schema Validation

Prebuilt Workflows

Quick Start

Connect to MCP Moira in under 2 minutes

claude.ai - Browser chat (most popular)

Settings → Connectors

# On claude.ai (Pro/Max/Team/Enterprise):

# 1. Go to Settings → Connectors
# 2. Click "Add custom connector"
# 3. Enter:
#    Server URL: https://moira-dev.with-andrei.com/mcp
# 4. Click "Connect"
# 5. Complete OAuth authentication in browser
# 6. Tools appear in chat

# Requires paid plan

chat.openai.com - Browser chat

Settings → Connectors

# On chat.openai.com (Plus/Pro required):

# 1. Profile → Settings
# 2. Go to "Connectors" or "Integrations"
# 3. Click "Add connector"
# 4. Enter:
#    Name: MCP Moira
#    URL: https://moira-dev.with-andrei.com/mcp
# 5. Complete OAuth authentication
# 6. Tools available in chat

# Free tier doesn't support MCP

Recommended: Use CLI command

Terminal

claude mcp add --transport http moira https://moira-dev.with-andrei.com/mcp

Then authenticate:

OAuth Flow

# After adding, authenticate within claude
/mcp
# → Select "moira"
# → Click "Authenticate"
# → Browser opens for OAuth

Alternative: Manual JSON config

# Alternative: Manual JSON config
# ~/.config/claude/mcp.json
{
  "mcpServers": {
    "moira": {
      "type": "http",
      "url": "https://moira-dev.with-andrei.com/mcp"
    }
  }
}

# Then: /mcp → Authenticate

Desktop app: GUI method

Settings → Connectors

# In Claude Desktop app:

# 1. Open Settings (⌘+,)
# 2. Go to "Connectors" tab
# 3. Click "Add custom connector"
# 4. Enter:
#    Server URL: https://moira-dev.with-andrei.com/mcp
# 5. Click "Connect"
# 6. Browser opens → OAuth → Done

# No file editing

Recommended: One-click install

Add to Cursor

Then authenticate:

OAuth Flow

# After clicking the button:
# 1. Cursor opens with install prompt
# 2. Click "Install" to add moira MCP server
# 3. Settings → MCP Servers → Find "moira"
# 4. Click "Authenticate" → Browser opens for OAuth

Alternative: Manual JSON config

~/.cursor/mcp.json

# ~/.cursor/mcp.json
{
  "mcpServers": {
    "moira": {
      "url": "https://moira-dev.with-andrei.com/mcp"
    }
  }
}

Recommended: One-click install

Add to VS Code

Then authenticate:

OAuth Flow

# After clicking the button:
# 1. VS Code opens with install prompt
# 2. Click "Install" to add moira MCP server
# 3. Settings → MCP Servers → Find "moira"
# 4. Click "Authenticate" → Browser opens for OAuth

Alternative: Manual configuration

settings.json

# Install MCP extension:
# ext install mcp-connector

# Then in settings.json:
{
  "mcp.servers": {
    "moira": {
      "url": "https://moira-dev.with-andrei.com/mcp",
      "transport": "http"
    }
  }
}

# Or: Command Palette → "MCP: Add Server"

Mac app: With helper

Settings → Connectors

# Perplexity Mac App:

# 1. Install PerplexityXPC helper first:
#    Settings → Connectors → Install Helper
# 2. Click "Add Connector"
# 3. Enter:
#    Server Name: moira
#    Command: npx
#    Args: -y mcp-remote https://moira-dev.with-andrei.com/mcp
# 4. Complete OAuth authentication
# 5. Ask Perplexity to use MCP Moira tools

# Paid plan recommended

VS Code extension: Open-source AI assistant

config.yaml

# Continue extension in VS Code:

# 1. Install Continue extension
# 2. Open config: Ctrl+Shift+P → "Continue: Open config"
# 3. Add to config.yaml:
#
# mcp:
#   servers:
#     moira:
#       url: https://moira-dev.with-andrei.com/mcp
#       transport: http
#
# 4. Restart VS Code
# 5. Authenticate when prompted

Zed Editor: Fast code editor

settings.json

# Zed Editor:

# Edit ~/.config/zed/settings.json:
{
  "mcp": {
    "servers": {
      "moira": {
        "url": "https://moira-dev.with-andrei.com/mcp",
        "transport": "http"
      }
    }
  }
}

# Restart Zed → Authenticate → Tools ready

Gemini CLI (experimental):

~/.gemini/settings.json

# Gemini CLI (experimental):

# ~/.gemini/settings.json
{
  "mcp_servers": {
    "moira": {
      "url": "https://moira-dev.with-andrei.com/mcp",
      "transport": "http"
    }
  }
}

# After saving:
# gemini auth → OAuth

What Happens Next?

1.

Simply type in your AI chat: "Start user onboarding flow"

2.

The agent will guide you through an interactive onboarding that explains Moira by demonstrating a workflow in action

3.

You'll learn about available workflows and get started with your first real task!

Who Benefits?

See how MCP Moira transforms AI workflows

📋 Task Breakdown with Validation

Complex task? AI breaks it down, executes each step, and proves completion before moving forward.

Problem:

Solution:

Example Workflow:

IN

"Set up CI/CD pipeline"

Complete CI/CD with all proofs verified

Result: Final report with all evidence - every step validated before proceeding

How It Works

Four simple steps to controlled AI workflows

Connect MCP Client
Add MCP Moira server URL to your Claude Code, Cursor, or any MCP-compatible tool. One-time setup.
Start Workflow
Call start_workflow with your workflow ID. MCP Moira returns the first step with instructions and schema.
Execute Steps
Agent completes each step and calls execute_step. Response is validated against JSON Schema before proceeding.
Get Results
Workflow completes with all data validated and context preserved. Every step executed in order.

Common Workflows

Pre-built workflows for typical AI agent tasks

Software Development

7 steps

Full development cycle: requirements analysis, architecture design, implementation, testing, and deployment with validation at each stage.

Product Selection

5 steps

Market research, feature comparison, price analysis, and recommendation generation for purchasing decisions.

Content Creation

5 steps

Topic research, outline planning, draft writing, editing, and SEO optimization with quality gates.

Research Pipeline

4 steps

Structured research process with source validation, competitor analysis, and report generation.

Code Review

5 steps

Automated code review with security, performance, and quality checks validated before proceeding.

Data Analysis

6 steps

Data validation, transformation, statistical analysis, and visualization with quality checks.

MCP Protocol Architecture

How MCP Moira orchestrates AI agent workflows

Claude Code

Cursor

VS Code

Any MCP Client

MCP Protocol

MCP Moira Server

moira-dev.with-andrei.com/mcp

Workflow Engine

Sequential execution

Validator

JSON Schema check

State Store

Context persistence

Validated Response

Controlled Workflow Execution

Before & After

See the difference MCP Moira makes in AI workflow control

Without MCP Moira

// Agent can skip steps
agent: I'll do task A, B, C

// No validation
agent: Here's the result
result: { data: "maybe correct?" }

// Lost context
agent: What was step 2 about?

// Unpredictable format
error: Cannot parse response

Steps can be skipped or reordered
No guarantee on response format
Context gets lost between steps
Manual retry on failures

With MCP Moira

// Forced sequential execution
step 1: Complete task A
✓ validated

// JSON Schema validation
step 2: { "status": "ok", "data": [..] }
✓ schema valid

// Automatic state management
step 3: Using context from step 1-2
✓ context preserved

Strict step-by-step execution
Guaranteed data structures
Persistent context across steps
Auto-retry with clear errors

JSON Schema Validation

Define the exact structure you need, MCP Moira enforces it

Schema Definition

{
  "type": "object",
  "properties": {
    "analysis": {
      "type": "string",
      "minLength": 50
    },
    "confidence": {
      "type": "number",
      "minimum": 0,
      "maximum": 100
    },
    "recommendations": {
      "type": "array",
      "items": { "type": "string" },
      "minItems": 1
    }
  },
  "required": ["analysis", "confidence", "recommendations"]
}

Validation Result

Valid Response

{
  "analysis": "The codebase uses...",
  "confidence": 85,
  "recommendations": [
    "Add error handling",
    "Improve test coverage"
  ]
}

Invalid - Auto Retry

{
  "analysis": "Short", // minLength: 50
  "confidence": 150    // max: 100
  // missing: recommendations
}

Auto Retry

Invalid responses trigger automatic retry with clear error message

Type Safety

Guaranteed data types for every field in the response

Required Fields

Ensure all critical data is present before proceeding

What is MCP Moira?

Workflow orchestration engine powered by MCP Protocol

MCP Moira connects to any tool with MCP support in just a couple of clicks:

• Claude Code ✓
• Cursor ✓
• VS Code (via MCP extensions) ✓
• Your custom tool ✓

One server — multiple clients.

Sequential Control

Force step-by-step execution with control at each stage

Response Validation

JSON Schema validation before proceeding to next step

State Management

Automatic context tracking between steps

Universal Integration

Works with Claude Code, Cursor, VS Code and any MCP client

Why Control Matters

Core capabilities that give you full control over AI agent behavior

Step-by-Step Control

Agent must complete each step properly before continuing. No uncontrolled execution.

Response Validation

JSON Schema enforcement ensures agents return valid data structures every time.

State Management

Automatic context persistence between steps. State never gets lost.

Retry Logic

Auto-retry failed validations with clear error messages to agent.

MCP Native

Works with any MCP client. Not locked into one tool.

Type Safety

Guaranteed data structures through JSON Schema validation.

Workflow Notifications

Get notified when workflows complete. Telegram integration for real-time updates.

Ready-to-Use Templates

Pre-built workflow templates. Agent handles technical details - you focus on results.

Product Roadmap

Available Now

MCP server integration, Step-by-step workflow control, JSON Schema validation, State management, Claude Code support

In Development

Visual workflow designer (Web UI), More workflow templates, VS Code MCP extension support, Advanced analytics

Planned

Workflow marketplace, Team collaboration features, Workflow versioning

Ready to Control Your AI Workflows?

Connect MCP Moira to your favorite tool and start building structured workflows today.

Get Started

View on GitHub

Quick Start

What Happens Next?

Who Benefits?

📋 Task Breakdown with Validation

💻 Code Review Automation

📊 Report Generation Pipeline

🧪 Test Case Generation

📣 Marketing Campaign Automation

🔬 Research Pipeline Automation

📅 Sprint Planning Automation

⚙️ Daily Workflow Automation

How It Works

Connect MCP Client

Start Workflow

Execute Steps

Get Results

Common Workflows

Software Development

Product Selection

Content Creation

Research Pipeline

Code Review

Data Analysis

MCP Protocol Architecture

Before & After

JSON Schema Validation

What is MCP Moira?

Sequential Control

Response Validation

State Management

Universal Integration

Why Control Matters

Step-by-Step Control

Response Validation

State Management

Retry Logic

MCP Native

Type Safety

Workflow Notifications

Ready-to-Use Templates

Product Roadmap

Available Now

In Development

Planned

Ready to Control Your AI Workflows?

Beta Version - Active Development