Skip to content

Examples & starter templates

Three ways in, shortest first.

MCP and skills (Claude Code, Cursor, any MCP client)

Section titled “MCP and skills (Claude Code, Cursor, any MCP client)”
/plugin marketplace add candledottv/agentic

That line (Claude Code shown; the repo has install docs for Cursor, Codex, OpenCode, and Grok Build) gives your agent five Candle skills. Add the MCP server to your client’s config and it gets launch, trade, swap, transfer, market, forensics, and activity tools without writing HTTP. If you have the CLI, candle mcp --print-config prints that config with no key to paste; otherwise the server is on npm as @candledottv/mcp. Reference: Candle MCP server, Skills for coding agents.

import { CandleClient } from "@candledottv/agent-sdk"
const candle = new CandleClient({ apiKey: process.env.CANDLE_API_KEY })
// Launch (idempotent: the SDK generates a clientLaunchId and retries safely)
const launch = await candle.launch({
chain: "solana",
name: "GLOW",
symbol: "GLOW",
imageUrl: "https://.../glow.png",
})

Reference: TypeScript SDK.

Terminal window
# Dry-run a launch with a test key
curl -X POST https://staging.api.candle.tv/api/v1/launch/headless \
-H "x-api-key: cndl_test_..." \
-H "Content-Type: application/json" \
-d '{"name":"GLOW","symbol":"GLOW","imageUrl":"https://.../glow.png","dryRun":true}'

Reference: Headless launch API, Agent trading API, Market state API.

  1. Store the key outside your code (CANDLE_API_KEY); it is shown once at creation.
  2. Start on a cndl_test_ key and dryRun until your happy path works.
  3. Subscribe to Webhooks instead of polling where you can.
  4. Set spend limits before you hand the loop to the agent: Wallets, delegation & custody.