Skip to content

Quick Start

This takes you from a clone to an agent answering in the console, all on your machine.

  • Node.js 22 (the repository pins it in .nvmrc) and git.
  • Nothing else for the local runtime: Cortex downloads uv, Python 3.12 and OpenTofu itself.
  • A model to call. The new-agent form defaults to Ollama on http://localhost:11434; any OpenAI-compatible endpoint works too.
Terminal window
git clone https://github.com/monaccode/astromesh-cortex.git
cd astromesh-cortex
npm install
npm run dev # Electron with hot reload

npm run build compiles the app into out/ and npm test runs the test suite.

  1. Create a workspace. On the Welcome page choose New Workspace…, pick a folder and a name. Cortex writes .cortex/workspace.json (commit it) and .cortex/local.json, which it adds to .gitignore. Open Folder… turns an existing folder into a workspace.

  2. Install the local runtime. Open the Runtime panel in the activity bar and click Install & Start. Cortex downloads uv, installs Python 3.12, creates a virtual environment, installs astromesh[mesh], astromesh-cli and astromesh-orbit[gcp], downloads OpenTofu, checks the import, and starts the runtime on 127.0.0.1:8000. It needs no Python of your own.

  3. Point Cortex at it. In Settings → Runtime Connections, add a connection of type local with URL http://localhost:8000 and click it to make it active. The status bar turns green.

  4. Create an agent. In Agents, choose New Agent → Blank Agent, fill in a name and a model, and create it. It opens in the YAML editor. From Template… builds one from the runtime’s templates instead.

  5. Deploy it. Click Deploy in the editor toolbar. The agent’s status turns deployed.

  6. Talk to it. Click Test. The console opens on the right; send a message and watch the tool calls and the model’s reasoning arrive live. Click View Trace on an answer to see every span.