Eliza — TypeScript AI Social Agent Framework

What it is

Eliza is a TypeScript framework for building autonomous AI agents that operate across social platforms. It supports Discord, Telegram, Twitter, and Slack as communication channels. Agents are defined through character files that specify personality, goals, and behavior. The framework includes a plugin ecosystem for extending agent capabilities and RAG-based memory for context retention.

Eliza targets developers building social AI agents, community bots, and autonomous assistants. The character-driven approach means you define who the agent is and how it behaves, not just what it does.

How it saves time or tokens

Eliza provides the platform integration layer that you would otherwise build from scratch. Discord OAuth, Twitter API, Telegram Bot API, and Slack integration are handled by the framework. The character file system means you configure agent behavior declaratively instead of writing custom handler code. RAG memory gives agents long-term context without manual state management.

How to use

1. Create a new agent project: npx elizaos create my-agent.
2. Copy .env.example to .env and add your API keys and platform tokens.
3. Define your agent's character in character.json and start with npx elizaos start.

Example

# Create and start an agent
npx elizaos create my-agent
cd my-agent
cp .env.example .env
# Edit .env with API keys
npx elizaos start

{
  "name": "TechBot",
  "bio": "A helpful AI assistant for developer communities",
  "lore": ["Passionate about open-source software"],
  "messageExamples": [
    [
      {"user": "user1", "content": "What is TypeScript?"},
      {"user": "TechBot", "content": "TypeScript adds static types to JavaScript. It catches bugs at compile time."}
    ]
  ],
  "plugins": ["@elizaos/plugin-discord", "@elizaos/plugin-twitter"]
}

Related on TokRepo

• Agent Tools — AI agent frameworks
• Automation Tools — Bot and automation platforms

Common pitfalls

• Each platform has rate limits and terms of service. Aggressive posting or messaging on Twitter or Discord can get your agent banned. Configure posting frequency conservatively.
• Character files define agent personality but do not prevent all undesirable outputs. Add explicit constraints in the character's rules and lore sections.
• RAG memory grows over time. Implement memory pruning or summarization to prevent context windows from filling with irrelevant historical data.