Using MCP With Concludia

MCP lets you use an LLM model to work with Concludia as an argument workspace. Instead of copying an argument into a chat window as prose, you can let the MCP client read the argument's structure: conclusions, premises, values, counterpoints, proof state, and the actions your credential is allowed to take.

That makes MCP useful for two broad tasks: creating new arguments and improving arguments that already exist.

Creating Arguments

Concludia arguments can be easier to build with a collaborator. An MCP client can help turn notes, outlines, or rough prose into a first structured draft.

For example, you might ask it to:

• identify the main conclusion in a paragraph;
• separate factual premises from value premises;
• point out where the reasoning depends on an unstated assumption;
• validate the draft before you import it into Concludia.

This is useful because the first version of an argument is often messy. MCP can help you get from "I think this is my case" to a graph that can be reviewed, contested, and improved.

Reviewing Arguments

MCP can also help when an argument is already in Concludia. A client can list the graphs you can access, read a whole argument, or focus on a smaller graph slice around a particular conclusion.

For review work, you might ask it to:

• summarize what the argument is trying to prove;
• find premises that seem weak, ambiguous, or too broad;
• suggest possible counterpoints;
• explain which parts of the graph appear stable and which deserve closer attention.

Using MCP to review can make the structure easier to inspect so you can decide where your attention should go.

Adding Counterpoints

MCP is especially helpful when you want to respond to someone else's argument. If a public graph accepts counterpoints, a client can help you draft a focused objection.

A good prompt is explicit about the boundary:

Review this argument and draft one counterpoint I could add. Before suggesting any contribution, check what actions I am allowed to take on the graph.

That boundary matters because responding to someone else's argument is not the same thing as rewriting it. Currently, authors can edit their own graphs. Other readers may be allowed to add counterpoints, but they cannot add premises, change statements, or change the support relationships in the author's graph. In that situation, the right action is to add a counterpoint that clearly identifies the problem you see.

Working With Permissions

Concludia exposes graph-specific capabilities to MCP clients. A client can see whether your current credential can read a graph, add counterpoints, edit the graph structure, or manage graph-level settings.

This helps keep suggestions grounded in what you can actually do. If you are reviewing a graph where only counterpoints are allowed, the client should steer you toward counterpoints. If you are using a read-only credential, it should help you review without attempting mutations.

When in doubt, ask the client to check capabilities first.

Connecting A Client

How you connect depends on the client you are using. Some MCP-capable clients offer a Concludia sign-in flow. Others ask you to paste connection details from your Concludia profile.

If your client can sign in to Concludia directly, follow that flow and review the access it is asking for. If your client asks for manual connection details, use this reference:

• Open your Concludia profile.
• Create an MCP developer token.
• Copy the MCP endpoint and the token shown after creation.
• Paste them into your client's MCP connection settings.
• Choose the narrowest access that fits the work you want to do.

Use read-only access when you want the client to inspect arguments without changing anything. Use read-write access when you want it to help create graphs, import drafts, or add allowed contributions.

Tokens are shown once. If you lose a token or no longer trust it, revoke it from your profile and create a new one.

Useful First Prompts

After a client is connected, useful first prompts include:

• "List my Concludia graphs and tell me what I can do with each one."
• "Help me turn these notes into a Concludia argument draft."
• "Review this argument for weak premises and possible counterpoints. Do not mutate anything."
• "Draft a counterpoint I could add to this public argument, but first check the graph capabilities."

MCP works best when you treat the client as a reasoning assistant, not as an authority. Let it help with structure, review, and drafting, then decide what you actually want to accept or add.