Add AI to your legacy & enterprise systems
without a rewrite

legacy2mcp auto-generates Model Context Protocol (MCP) servers from WSDL/SOAP services (databases and queues next), with safe defaults so AI agents can use old systems without putting them at risk.

★ Star on GitHub Read the blog
Open source · Apache 2.0 · Python

Downloads on PyPI

–
Last day
–
Last 7 days
–
Last 30 days
–
All time

Loading live stats from PyPI…

AI transformation for legacy and enterprise systems

Most AI transformation projects stall on the same problem: the data and business logic that matter live in legacy and enterprise systems, such as SOAP web services, ERP and CRM back ends, mainframe integration layers and core banking or insurance platforms, and AI agents can't reach them. Rewriting or migrating those systems first takes years.

legacy2mcp takes the other route. It connects legacy systems to AI by generating a Model Context Protocol server in front of what you already run, so you can add AI capabilities to legacy systems in days instead of replacing them.

Legacy system to AI

Turn an existing WSDL into tools that Claude, ChatGPT and other MCP-compatible AI agents can discover and call.

Enterprise system to AI

Give agents governed access to systems of record, with read-only defaults, schema validation and audit logs.

Modernize without migrating

Legacy modernization with AI that leaves the original system untouched, so there's no rewrite and no risky cut-over.

Why legacy2mcp

Generated, not hand-written

Point it at a WSDL and get an MCP server whose tools mirror the service's operations.

Safe by default

No arbitrary SQL and no writes unless you explicitly opt in.

Validated inputs

Every tool call is checked against a JSON Schema before it reaches the legacy system.

Audit logging

See what an agent called, with what arguments, and when.

Pluggable adapters

A small adapter interface makes it possible to add new system types.

Apache 2.0

Free to use, modify and embed in commercial work.

Get started

pip install legacy2mcp
# then follow the README for usage

Full documentation lives in the GitHub repository.

Roadmap

AdapterStatus
WSDL / SOAPImplemented
Databases (Postgres, SQL Server, Oracle)Planned
Queues (Kafka, RabbitMQ, SQS)Planned

Guides

Strategy

AI transformation for legacy systems

How to add AI capabilities to enterprise systems without a rewrite.

Concepts

What is MCP, and why do legacy systems need it?

The problem MCP solves for AI agents and old enterprise services.

How-to

How to connect a SOAP/WSDL service to an MCP server

Mapping operations to tools, step by step.

Security

MCP server security: safe defaults

Guardrails for agents touching systems of record.

Comparison

MCP vs REST APIs for AI agents

When each one is the right tool.

Architecture

The adapter pattern for connecting legacy systems to agents

A design that scales past the first integration.

FAQ

What is legacy2mcp?

An open-source, Apache 2.0 Python toolkit that generates MCP servers from legacy systems, starting with WSDL/SOAP services.

Which systems are supported?

WSDL/SOAP first. Database and queue adapters are planned.

Is it safe to use?

It defaults to safe behavior: no arbitrary SQL or writes without explicit opt-in, JSON Schema validation, and audit logging.

How do I add AI capabilities to a legacy system?

Put an MCP server in front of it. legacy2mcp generates one from the system's existing interface (a WSDL today), so AI agents can call its operations as tools with no change to the legacy system.

Can it connect ERP, mainframe or other enterprise systems to AI?

Yes, if the system exposes a SOAP/WSDL web service, which many ERP, CRM, banking and mainframe integration layers do. Direct database and queue adapters are planned.

Is this AI transformation without replacing legacy systems?

Yes. Instead of migrating a working system of record, you expose it to AI agents through MCP, with safe defaults.

Can I add my own adapter?

Yes, through the adapter interface.