Connect your agentComing soon

Your agent calls Mondian’s decision tools. Keep your agent and your model.

Connect Claude, Codex or any agent that supports MCP. Each decision is calculated from your data and rules, and comes back with its record.

Book a demo See the decisions

In Claude Code, add Mondian as a server with one command.
In the Claude app, open Settings → Connectors, add a custom connector and paste the address.

claude mcp add --transport http mondian https://your-mondian-address

Add Mondian to your Codex configuration.
Start Codex; the decision tools are listed with its other tools.

# ~/.codex/config.toml
[mcp_servers.mondian]
url = "https://your-mondian-address"

In ChatGPT, add a custom connector.
Paste your Mondian address.

https://your-mondian-address

Add Mondian to your project's MCP file.
Cursor lists the decision tools under MCP.

// .cursor/mcp.json
{
  "mcpServers": {
    "mondian": { "url": "https://your-mondian-address" }
  }
}

Any agent or framework that supports MCP over HTTP can connect.
Add Mondian as a remote server with your address.

// remote MCP server
{ "name": "mondian", "transport": "http", "url": "https://your-mondian-address" }

Your agent signs in to your Mondian workspace. It sees only the decisions and data that workspace allows.

Setup

Three steps. Then every call works the same way.

01Sign up
Create your Mondian workspace. You get the address your agent connects to.
02Add Mondian to your agent
One address in Claude, Codex, ChatGPT, Cursor or your own agent.
03Call a decision tool
Your agent sees every decision your workspace has, with its inputs and what it returns.

A call

Your agent asks for a decision. It gets the decision and its record.

Tool call

One call, one decision.

Your agent passes the inputs. Mondian calculates the decision from your data and applies your rules.

// your agent callsreorder_stock({ "category": "Garden furniture", "week": 43 }) // Mondian returns{ "decision": "117 stores, 41,600 units", "status": "waiting_for_approval", "reason": "38 stores are above 500 units", "record": "rec_…"}

Record

Every call writes a record.

What was decided, from which data, under which rules, and who approves it.

Decision recordExample

Reorder stock · Garden furniture · week 43

Made by: Your agent · Reorder stock
Approval: Waiting for Category planner
Data: sales_weekly v4 · stock 06:00
Rules: Approval above 500 units per store

Why a tool

Your agent stops improvising.

Without a tool, the model works the number out again on every run. A Mondian decision tool runs as code, so the same inputs give the same result.

Same question, three runsExample

Units to reorder · Store 12 · garden furniture · week 43

Prompt only380520430

Mondian tool410410410

See the decisions your agent can call

Questions

Before you connect.

Which agents can connect?

Any agent that supports MCP, including Claude, Codex, ChatGPT and Cursor, and agents you build yourself.

Do we still need the Mondian agent?

No. Use your own agent, the Mondian agent, or both. They call the same decision tools and write the same records.

Isn't a skill enough?

A skill is text the model reads, and it may or may not follow it. A Mondian decision tool runs outside the model, so the calculation and the rules are applied every time.

Which decisions can my agent call?

Every decision in your workspace: the ones Mondian has defined and the ones your team adds.

Where does our data go?

Your data is never used to train models. It is processed in an isolated sandbox and deleted afterwards, or Mondian runs inside your own systems.