Paid tier

Connect your AI

Describe a site to your own AI assistant and get a solved Levelised model back in your cloud projects.

The assistant in the app builds models from a conversation on every tier. On the paid tier you can also do it from the AI assistant you already use, without opening the app until the model is built.

Levelised exposes its model builder as an MCP server, the open protocol that lets an assistant use tools. Connect Claude Code or Cursor to it, then talk to the assistant the way you would to a colleague:

Build a microgrid in Levelised for a mine in the Kimberley with a 100 MW load. Solar but no wind. There is no gas pipeline, so firm supply is trucked diesel or trucked LNG. Compare the two.

The assistant asks a couple of questions it cannot default, the delivered fuel prices in this case, because nothing published settles those for your site. Then it works through the model in the order you would: the buses, the load, each supply option with its cost from CSIRO GenCost and its emission factor from the National Greenhouse Accounts Factors, a solar trace for the location from the same data the map picker uses (about 35% single-axis at Fitzroy Crossing; gross, from PVGIS, so a screening number), then validate, solve the full year and tell you the levelised cost with a source against every figure. Say yes and the model appears under “Open from Cloud” in the app, solved, laid out and coloured by carrier. Or ask for a download link and open the file with File → Open.

Set-up

  1. Open Levelised, sign in, and choose “Connect your AI…” from the account menu. Create a key and copy it; it is shown once.
  2. Add the server to your assistant. Claude Code, in a terminal:
    claude mcp add --transport http levelised https://app.levelised.com/mcp \
      --header "Authorization: Bearer YOUR_KEY"
    Cursor: add an MCP server with the URL https://app.levelised.com/mcp and the request header Authorization: Bearer YOUR_KEY.
  3. Say “Levelised” in your request. The assistant reads the server's modelling playbook before its first build; you do not need a special prompt. If you want to see what it can do first, ask it “what can I do with Levelised?”

Claude Desktop and claude.ai sign in to connectors in a different way, which Levelised does not support yet. If you'd like a hand with any of this, tell us and we'll set it up with you.

What it will and won't do

The server's playbook tells your assistant never to make up a number and gives it the tools to look one up. Technology costs, efficiencies and lives come from GenCost, emission factors from the NGA Factors, and anything site-specific comes from you or from a source the assistant names and records in the model. Hourly time series never pass through the conversation; they stay in Levelised. It writes only the model it built, under the name you agree, and never touches your other projects.

Your tier's limits apply as in the app, and the solves your assistant runs count towards them. When one is reached the assistant tells you. Your own assistant does not draw on the in-app assistant's weekly allowance. What it costs to run is between you and its supplier.

A key acts as you, so keep it out of shared files and revoke it from the same dialog if it may have been exposed.

Try it

Create a key, connect Claude Code, and ask for the Kimberley microgrid above. Then open the result in the app, click the gensets, and check the efficiency and cost basis against what the assistant told you. That is the check we run on every build, and it is the right habit.

Back to the documentation.