Solonic

Workflow Based MCP Servers

Documentation
What are Workflow-Based MCP Servers?

We're building an application which allows businesses to create workflow-based MCP servers from their already existing APIs.

The Problem with API-Driven MCP Servers

Current MCP server generators turn APIs into MCP tools on a 1:1 mapping; each API becomes its own tool. We consider these MCP servers to be API driven, since they're created based on the business' already existing API definition.

Example: Meeting Creation

When a user asks an agent to "book a meeting with John at 10:00 AM on 7/20", an API-driven MCP server might require:

  1. Call a check conflicts tool to ensure John is free
  2. Call an add to calendar tool to add the meeting
  3. Call a send meeting invite tool to notify John
Performance and Cost Issues

Each additional sequential tool call causes performance degradation and increased costs:

  • Performance degradation: Each tool call requires LLM inference, increasing latency
  • Accuracy loss: LLMs perform less accurately as sequential tool calls increase
  • Higher costs: Increased token usage from multiple API calls
Our Solution: Workflow-Based MCP Servers

Our product breaks away from API-driven MCP servers by allowing users to create workflow-based MCP servers. We define workflow-based as creating MCP servers with tools that match the workflows of the users who will be using them.

Improved Example: Single Tool Call

Instead of requiring three consecutive tool calls, our product allows businesses to combine the APIs behind the meeting creation process into a single workflow-based tool. When a user requests to create a meeting with John, it only requires a single tool call.

How It Works
  1. Upload OpenAPI schemas: Import your existing API definitions
  2. Combine APIs: Group related endpoints into workflow-based tools
  3. Automated processing: Our internal agent lines up inputs and outputs
  4. Testing: Automatically test the combined tools
  5. Deployment: Deploy a remote MCP server with authentication and scaling
Benefits

Decreased Latency

Fewer tool calls mean faster response times

Lower Token Usage

Reduced API calls lead to cost savings

Improved Accuracy

Single tool calls perform more reliably

Getting Started
Ready to transform your APIs into workflow-based MCP servers? Here's how to get started:
  1. Sign in with your Google account
  2. Upload your OpenAPI schema files
  3. Select the endpoints you want to include
  4. Create workflow-based tools that match user workflows
  5. Configure security credentials
  6. Generate and deploy your MCP server