Run Your Own

The /goal runtime is open source and deliberately small — a few hundred lines you can read in one sitting. It gives any token an autonomous operator: market observation, LLM reasoning, action queue, and persistent memory, in one loop.

Quickstart

git clone https://github.com/BethorDevelop/goal.git
cd goal
npm install

# point the loop at your token
npx goal run \
  --mint <your-pump-fun-mint> \
  --objective "reach a $1,000,000 market cap" \
  --interval 150

Plugging in a model

The runtime ships with a provider interface, not a vendor lock. Set one environment variable and implement one function:

// src/llm.ts
export interface LlmProvider {
  complete(prompt: string, context: string[]): Promise<string>;
}

// wire any provider: Fable 5, or anything with an API
export LLM_API_KEY=...   // your provider key

Without a key, the loop runs in heuristic mode — useful for watching the mechanics (cycles, memory, market reads) before spending a single credit.

What it does out of the box

• Market observation via DexScreener's public API — price, market cap, volume, liquidity for any mint.
• Goal tracking — progress computation, gap analysis, and a JSONL ledger of every snapshot.
• Reasoning cycles — your provider receives the snapshot plus memory and returns a mission with actions.
• Persistent memory — append-only JSONL with periodic consolidation, so the loop compounds across restarts.
• A local log — the same event-stream format that powers Mission Control, in your terminal.

What you bring

Publishing integrations (X, Discord, Telegram) are intentionally left as adapters for you to wire — every platform has its own keys, ToS and rate limits, and you should own that decision. The repo includes the adapter interface and an example.