EDI Enters The Coding Agent Era

Tediware has launched an official Model Context Protocol (MCP) server for EDI. This unlocks significant productivity gains for developers, EDI professionals and hands-on business operators, and makes EDI far more accessible and easier to use for everyone.

You can try it out in your coding agent right now: just point it at the MCP documentation. You will need an API key, so if you haven’t signed up for Tediware yet, you can sign up free. Or, watch the demo video.

Some of the things you can do

• Easily connect with your customers and suppliers. Adding EDI to your applications, especially your own custom apps, can be done in minutes. What used to take days or weeks of specialist time is a snap with Claude Code or Cursor.
• Get fast, thorough answers about your data and your business. Instead of digging through logs and reports, talk to your agent. It has all the tools it needs to find, analyze and report on your data.
• Deliver better support. When things go wrong in EDI, as they often do, it used to be time-consuming and frustrating to figure out what happened and how to fix it. It doesn’t have to be that way, and with Tediware, it isn’t.
• Create new, innovative apps that use EDI. Build an EDI-native TMS, accounting system or ERP by ingesting inbound EDI documents, turning them into your data schema, and then adding outbound EDI functionality with a few prompts.

Built for humans

Tediware is not built to replace EDI professionals, it’s built to make them more productive and more valuable.

This is what has happened with software engineering, where the demand for experienced software engineers has grown, not declined, as coding agents become more capable. The same revolution in programming productivity is going to happen with EDI. Why hasn’t it happened already?

Some of the reasons are structural and have to do with the way some traditional EDI companies operate, a topic I’ll save for another day. But there are also many technical reasons for why EDI integration work has lagged behind programming.

One is that traditional EDI software is visual. People create mappings by navigating complex UIs and drawing lines between boxes. Although Tediware’s AI-assisted mapping feature creates mappings in minutes, it does so in the context of a traditional web-based UI. These visual UIs are not easily or quickly controlled by coding agents.

With the launch of Tediware’s MCP server, that’s changed. It lets agents read, understand and interact with your configuration and your data in a way that is fast, natural, secure and controlled.

Technology exists to serve humans, and Tediware is no different. It helps people build integrations and entirely new software in less time and with less pain. It helps companies deliver value to their clients more quickly, improving their margins and helping them grow.

A step forward in agentic EDI autonomy

This update expands Tediware’s capabilities along each of the three axes of the agentic EDI autonomy scale:

• Context: The ability to read your control plane configuration (including partners, connections, mappings and implementation guides) lets your agent understand your setup, trace documents through flows and answer your questions.
• Agency: Access to sending and receiving data means your agent can resend failed transactions, help you test document flows by generating and sending test transactions, find documents you’re looking for (“find PO 123456 and tell me when it arrived”), build custom integrations using test or production data as examples, and so much more.
• Trust: We have continued building on months of investment in observation and validation so that you can keep trusting the platform, and your agent, with your sensitive data. Access to the control plane is currently read-only. Secrets like passwords are never exposed via MCP, and you can obfuscate customer data using the tedi CLI. The next section includes more details.

Frequently asked questions

What is MCP?

Model Context Protocol (MCP) is an open standard for connecting AI agents to external tools and data. An MCP server publishes a set of tools, such as “list transactions” or “retrieve a partner’s configuration”. An agent like Claude Code or Cursor can then call those tools while it works on your behalf, instead of you copying and pasting data into your chat window.

How is MCP relevant to EDI?

Most EDI work happens in visual interfaces and specialist tools that agents can’t work with. For example, you find a failed transaction on one screen, then you need to check the error on a second screen, and then review your partner’s configuration on a third. Because these screens are displayed in a web browser (or worse, from the agent’s perspective: a desktop application), they are not easily (or safely) accessible by an agent.

MCP changes that by letting your agent read traffic, configuration, logs and so on. It can trace a document from the moment it arrived and explain why it failed. Once it understands the failure, it can help you fix it.

What can I build with it?

Anything that needs to exchange EDI with trading partners. Add inbound orders to your own app, or add the ability to send invoices and advance ship notices to it. Build a TMS that accepts load tenders, or an ERP that ingests purchase orders. Your agent can study real documents from your account and use them as examples while it writes the integration. With a standard key, it can also send test transactions to check its work.

Not everything is a build, either. A lot of the value is in day-to-day support: finding a document, explaining an error or resending something that failed.

Do I need to be a developer?

No. If you can describe what you want in plain language, an agent can look things up for you: “Did we receive Acme’s PO 123456?” or “Why did yesterday’s invoice to Globex fail?” Connecting the server takes a few minutes in a terminal or an AI desktop app. After that, you just ask your agent.

Which agents does it work with?

Any agent or app that supports MCP, including Claude Code, Cursor, Claude Desktop and VS Code. Most of these launch MCP servers locally, so the tedi CLI includes a bridge, tedi mcp serve, that connects them to Tediware. The MCP documentation has the setup steps. Simply giving your agent a link to the documentation is usually enough to get set up.

Do I have to run my EDI through Tediware?

The tools work with the documents and configuration in your Tediware account, so your agent can only trace, search and resend traffic that flows through Tediware. If you’re thinking trying out Tediware, you don’t have to move everything at once. Many customers start with one trading partner.

Which EDI standards does it support?

X12, the North American standard. The server covers every X12 release Tediware supports, and your agent can ask which ones those are.

How is this different from the API and the tedi CLI?

All three use the same API key, the same rate limits and the same service terms, and they cover the same ground. The API is for code you write and run yourself. The CLI is for you, working in a terminal. The MCP server is for your agent. Many people use all three.

Can my agent send a document to a real trading partner by accident?

Only if you give it a key that allows sending. A read-only key never sees the tools that send, receive or resend documents. Use a standard key only when a task needs one, and set your agent to ask before it runs a write tool. The MCP documentation shows how to keep a read-only connection on by default and turn on sending only when you need it.

Can my agent change my EDI control plane over MCP?

No. Your agent can read your EDI control plane (partners, connections, mappings and implementation guides) via MCP but can’t change it.

Where does my data go?

Anything a tool returns goes to the AI model your agent runs on. That includes the business data in your transactions. Your AI provider’s terms govern what happens to it there, including whether it’s used for training, so check them. To keep customer data out of the model entirely, scrub documents locally first with tedi edi obfuscate. Secrets like passwords are not served through MCP.

Is it free?

Yes. There is no additional charge for MCP – it is available on every plan, including the free Developer plan, which includes 800 transactions a month and needs no credit card. Your agent’s AI usage goes on your own bill with your AI provider.

Can I look up what an X12 segment or element means through the MCP server?

No. Tediware’s license for the X12 standard doesn’t allow it to serve reference content through MCP. The server’s tools work with your own data. You can look up the standard in Tediware’s own tools, such as the EDI Inspector and the tedi CLI.

Licensing and security

• X12 reference content is not served via MCP. Tediware licenses the X12 standard from X12 Incorporated, and that license permits AI use of X12 reference content only inside Tediware’s own features. The MCP server doesn’t serve reference content. Its tools work with your data only.
• One key per agent. Every MCP request authenticates with an API key from your dashboard, and a key’s scope is set when it’s created. A read-only key never sees the tools that send, receive or resend documents. The tools that change something are marked as write tools, so you can configure your agent to ask before running one.
• Same terms, same limits. MCP runs under the same service terms and rate limits as the API. A key works only after someone in your organization has accepted the terms.

Read our security page for more information on our approach to security.

Questions? Comments?

If you have questions or comments, drop me a line at adrian@tediware.com