Hackathon Quickstart

A single copy-paste reference for the facts you need in the first thirty minutes: where the MCP endpoint is, which chains are supported, which USDC address and faucet to use, what the two API key types are for, and the rate limits. Each section links to its full reference page.

1. Connect the MCP endpoint

The hosted MCP server is the fastest way to drive KeeperHub from an AI agent.

claude mcp add --transport http keeperhub https://app.keeperhub.com/mcp

Run /mcp in Claude Code to complete OAuth in the browser. For headless or CI environments, pass an organization API key instead:

claude mcp add --transport http keeperhub https://app.keeperhub.com/mcp \
  --header "Authorization: Bearer kh_your_key_here"

Every listed marketplace workflow is also reachable as its own typed MCP server at https://app.keeperhub.com/mcp/w/<slug>. See the MCP Server reference for the full tool list and per-workflow details.

The endpoint URL is also shown, with a copy button, in the dashboard under Settings -> API Keys.

2. Pick a key type

KeeperHub has two key systems. They are not interchangeable.

For programmatic API and MCP access, use an organization (kh_) key. Full details: API Keys.

3. Supported chains

Status reflects support maturity: stable chains are production-ready; experimental chains are accepted but may behave unreliably (for example, broadcasts can hang) and should not be used for production writes without opting in explicitly.

Start on a testnet. Fund the wallet with native gas first, then test USDC.

Testnets (recommended for hacking)

Mainnets

Experimental

The live source of truth for chains is GET /api/chains; agents can read the same list (including per-chain status) from the list_action_schemas MCP tool. Faucet links are third-party and may change. See the full Chains reference for chain-name aliases and ABI fetching.

4. Rate limits

Rate-limited requests return 429 with a Retry-After header. See Direct Execution for spending caps.

5. Sandbox

The Code action runs untrusted JavaScript in an isolated node:vm sandbox with outbound SSRF protection. See Code Plugin for what is allowed and blocked.