ElizaOS Expert Guide - January 2026

ElizaOS Expert Guide - January 2026

What is ElizaOS?

ElizaOS is a framework for building AI agents with:

• Plugin System - Modular actions, providers, evaluators, services
• Multi-Platform - Discord, Telegram, Twitter, Direct clients
• Character System - Personality, knowledge, examples
• Memory Management - Vector database, conversation history
• Model Flexibility - OpenAI, Anthropic, local models

Version: 1.7.0+ GhostSpeak Plugin: @ghostspeak/​plugin-elizaos

When to Use This Skill

• Building custom ElizaOS plugins
• Creating agent characters (like Caisper)
• Implementing custom actions or providers
• Integrating with GhostSpeak smart contracts
• Setting up Telegram/​Discord bots
• Managing agent memory and context

Core Concepts

Plugin Structure

import { Plugin } from "@elizaos/​core";

const myPlugin: Plugin = {
  name: "my-plugin",
  description: "My custom plugin",
  actions: [myAction],
  providers: [myProvider],
  evaluators: [myEvaluator],
  services: [myService],
};

export default myPlugin;

Actions (Agent Capabilities)

import { Action, HandlerCallback } from "@elizaos/​core";

export const myAction: Action = {
  name: "MY_ACTION",
  similes: ["DO_THING", "PERFORM_ACTION"],
  description: "Does something useful",

  validate: async (runtime, message) => {
    // Return true if action should be triggered
    return message.content.text.includes("do thing");
  },

  handler: async (runtime, message, state, options, callback) => {
    // Perform action
    const result = await doSomething();

    // Send response
    if (callback) {
      await callback({
        text: `Done! Result: ${result}`,
      });
    }

    return true;
  },

  examples: [
    [
      {
        user: "{{user1}}",
        content: { text: "Please do the thing" },
      },
      {
        user: "{{agentName}}",
        content: { text: "I'll do that right away!" },
      },
    ],
  ],
};

Providers (Data Sources)

import { Provider } from "@elizaos/​core";

export const myProvider: Provider = {
  get: async (runtime, message, state) => {
    // Fetch data
    const data = await fetchSomeData();

    // Return context to inject into prompts
    return `Relevant data: ${JSON.stringify(data)}`;
  },
};

Character Definition

import { Character } from "@elizaos/​core";

export const caisperCharacter: Character = {
  name: "Caisper",
  username: "caisper",
  bio: [
    "A friendly ghost AI that helps users navigate the GhostSpeak platform",
    "Expert in Solana, verifiable credentials, and reputation systems",
  ],
  lore: [
    "Born from the blockchain mists",
    "Guardian of agent reputations",
  ],
  messageExamples: [
    [
      {
        user: "{{user1}}",
        content: { text: "What's my Ghost Score?" },
      },
      {
        user: "Caisper",
        content: {
          text: "Let me check your reputation on-chain... Your Ghost Score is 750! That's pretty good!",
        },
      },
    ],
  ],
  postExamples: [],
  topics: ["solana", "blockchain", "reputation", "ai-agents"],
  adjectives: ["friendly", "helpful", "knowledgeable", "ethereal"],
  style: {
    all: ["Be friendly and helpful", "Use ghost emoji occasionally ={"],
    chat: ["Keep responses concise", "Be conversational"],
    post: ["Share insights about Web3 and AI"],
  },
  plugins: ["@ghostspeak/​plugin-elizaos"],
};

GhostSpeak Integration

Using @ghostspeak/​sdk in Actions

import { GhostSpeakClient } from "@ghostspeak/​sdk";

export const checkReputationAction: Action = {
  name: "CHECK_REPUTATION",

  handler: async (runtime, message, state) => {
    // Initialize SDK
    const client = new GhostSpeakClient({
      cluster: "devnet",
    });

    // Get user's wallet address from message
    const walletAddress = extractWalletAddress(message);

    // Fetch reputation
    const agent = await client.agents.getByOwner(walletAddress);

    return {
      text: `Your Ghost Score is ${agent.ghostScore}!`,
    };
  },
};

Telegram Integration (apps/​web)

// apps/​web/​lib/​telegram/​adapter.ts
import { TelegramClient } from "@elizaos/​client-telegram";
import { AgentRuntime } from "@elizaos/​core";

export async function setupTelegramBot(runtime: AgentRuntime) {
  const client = new TelegramClient(
    runtime,
    process.env.TELEGRAM_BOT_TOKEN!
  );

  await client.start();
}

Testing Plugins

import { describe, test, expect } from "bun:test";
import { AgentRuntime } from "@elizaos/​core";

describe("My Plugin", () => {
  test("action triggers correctly", async () => {
    const runtime = new AgentRuntime({
      token: "test",
      character: testCharacter,
    });

    const result = await myAction.validate(runtime, testMessage);
    expect(result).toBe(true);
  });
});

Development Workflow

# Install ElizaOS
pnpm add @elizaos/​core @elizaos/​client-telegram

# Create plugin
mkdir packages/​plugin-my-plugin
cd packages/​plugin-my-plugin
pnpm init

# Build plugin
pnpm build

# Test plugin
pnpm test

# Use in agent
import myPlugin from "@my-org/​plugin-my-plugin";
const runtime = new AgentRuntime({
  plugins: [myPlugin],
});

Best Practices

1. Validate inputs - Always validate user input in actions
2. Handle errors gracefully - Don't crash on API failures
3. Use providers for context - Keep actions simple, providers rich
4. Test thoroughly - Test all action paths
5. Document examples - Good examples improve agent behavior

Additional Resources

• Full plugin guide: /​pluginskill slash command
• ElizaOS Docs: https://github.com/elizaOS/eliza
• GhostSpeak Plugin: packages/​plugin-ghostspeak/
• Caisper Character: apps/​web/​server/​elizaos/​characters/​caisper.ts