MCP for tenders: Connect Auftrag One to AI assistants

An AI assistant can draft text and answer questions. It becomes much more useful for tender work when it can access the right data. That is why Auftrag One provides a server for the Model Context Protocol (MCP).

Through MCP, a compatible assistant can search tenders in Auftrag One, read notices and their documents, and move selected results into the bid pipeline. You no longer need to copy data between browser tabs, and the assistant works with your organization's tender data instead of an isolated chat history.

What is MCP?

The Model Context Protocol is an open interface that lets AI applications use external tools and data sources. Instead of building a separate integration for every assistant, Auftrag One makes its capabilities available through one standardized endpoint:

https://auftragone.com/mcp

This does not give the assistant unrestricted access to your workspace. It can only use the tools exposed by the Auftrag One MCP server, and it must authenticate through a user login or an organization API key.

What can the Auftrag One MCP server do?

The available tools cover three parts of everyday tender work.

Research tenders

The assistant can search the database using a free-text topic, fetch an individual notice, and find related procedures. A first question can become a structured research brief without moving search terms and results between applications.

For example:

Find current tenders for maintaining municipal charging infrastructure. Show the most relevant results with the contracting authority, deadline, and a short explanation. Do not change anything in my pipeline yet.

Review tender documents

For a notice, the assistant can list available document archives, read extracted text, and retrieve existing AI summaries. This supports an initial review of long specifications, eligibility requirements, and contract documents.

For example:

Read the available documents for this tender. Create an overview of mandatory requirements, evidence to submit, deadlines, and open questions. Reference the underlying document for every point.

Extracted text and summaries speed up a review, but they do not replace the authoritative source documents. Procurement and bid decisions should always be checked against the originals. Our tender requirements checklist offers a practical review sequence.

Maintain the bid pipeline

The MCP server can list saved tenders, bookmark a notice into the pipeline, and update its bid status. A conversation can therefore move beyond analysis and, after your decision, carry out the next workflow step.

Make write actions explicit:

First summarize why this tender may fit our profile. Only save it to the pipeline after I confirm.

This keeps research clearly separate from actions that change the team's shared work.

Set up Auftrag One in Claude Code

Claude Code needs a single terminal command:

claude mcp add --transport http auftragone https://auftragone.com/mcp

When the connection starts, Auftrag One opens the browser. Sign in, select your organization, and approve the connection. Claude Code can then discover the available Auftrag One tools and use them for relevant tasks.

Other MCP clients can use the same endpoint when they support remote MCP over Streamable HTTP and OAuth authentication. If a client does not support the browser flow, you can use an organization API key instead. The exact setup depends on the client.

See the Auftrag One MCP documentation for the complete, continuously updated instructions.

Three useful starting points for your team

MCP is most effective when it supports a clear, repeatable workflow rather than an overly broad task. These three starting points are easy to separate:

1. Research brief: The assistant searches a service area, summarizes the most relevant notices, and identifies deadlines.
2. Document check: The assistant reads available text and structures requirements, evidence, risks, and questions.
3. Pipeline preparation: The assistant explains its selection and moves confirmed notices into the bid pipeline.

Start with a read-oriented workflow and define when a person must confirm a change. Once the output fits your way of working, turn the workflow into a reusable instruction for your team. Our features overview shows how MCP connects with the rest of Auftrag One.

Grant access deliberately

An MCP client works with the permissions of the connected organization. Treat the connection like any other work tool that can access internal data:

• Connect only clients you trust.
• Never paste API keys into prompts or chat histories.
• Review what the assistant will change before approving a write action.
• Revoke connections that are no longer in use.

MCP does not make the decision for your team. It shortens the path from a question to a defensible decision and can record confirmed next steps directly in Auftrag One.

---

Ready to try the workflow? Open the step-by-step guide and connect your MCP client to Auftrag One.