Skip to main content
Each AI surface installs independently. Pick the one that matches what you are building, or install the MCP gateway first since it is the smallest step to a working tool call.
You can start without funds: MCP market reads need no wallet, and Mind has an offline example. When connecting a wallet, use test networks and keep seeds out of committed or shared files. dry_run is a KaleidoAgent setting, not a global MCP or Mind option.

​
MCP Servers

kaleido-mcp is the quickest path to a first tool call, published to npm. KALEIDO_NETWORK=signet points it at the signet KaleidoSwap API and Spark’s test network in one variable:
Directly launching the stdio server waits for an MCP client; it is not a chat prompt. Add the command to your client’s configuration, then request a tool call.
Spark and Liquid wallets are optional and not installed by plain npx. To enable them, run npx -y -p kaleido-mcp -p @tetherto/wdk-wallet-spark -p @kaleidorg/wdk-wallet-liquid kaleido-mcp and set WDK_SEED / LIQUID_MNEMONIC. For the RGB tools you also need an RGB Lightning Node. KaleidoCLI runs one on signet in Docker, see Build a Local RGB Agent for the full path. For the tool catalog and client configuration, see MCP Servers. The unified server includes the older wallet servers’ domains. Existing users can follow the migration notes.

​
KaleidoAgent

Prerequisites: Node.js 20 or newer, KaleidoCLI in $PATH, the Nanobot runtime, and an Anthropic or OpenAI API key.
To set the API key, configure the wallet in agent.config.json, start the dashboard, or deploy in a container, check the KaleidoAgent page.

​
KaleidoMind

KaleidoMind is an engine for your application. Start with its offline project template, use it through the Desktop App, or embed the package directly.

​
Try the offline starter

Expected result: sample balances, a mock invoice and a simulated transfer. This uses a fake wallet and scripted responses, with no node or model download. npm run start:mock adds a real local model while keeping the wallet simulated.

​
Through the Desktop App

The fastest path. The Desktop App bundles the engine as in-app chat, and manages the on-device model lifecycle. Switch it on from Settings > Capabilities by selecting Node + Mind or Only Mind.
KaleidoMind depends on a local AI runtime and may not be available on Windows yet. Every trading and node feature works without it, so run the app in Only Node mode if Mind is unavailable on your platform.

​
In your own Node.js app

@qvac/sdk is an optional peer dependency, so the engine also runs without it against a mock provider. The @kaleidorg/mind README has the quickstart, and the kaleido-mind repository carries runnable examples (examples/node-minimal, examples/rgb-agent).

​
From source, for development

The CLI in apps/cli handles on-device model management and runs the agent directly:
To exercise the engine against a model without a phone, use the playground from the repository root:
For how the tiered funnel routes a request, the tool contract shared across transports, the skills the engine loads, and where it runs on desktop and phone, check the KaleidoMind page.