Skip to content

Quick Start

Get an Automate-E agent running in 5 minutes. You'll start with test mode (no Discord needed), then optionally connect to Discord.

Prerequisites

Step 1: Clone and Install

git clone https://github.com/Stig-Johnny/automate-e.git
cd automate-e
npm install

Step 2: Create Your Agent

Create a file called character.json in the project root:

{
  "name": "My-Agent",
  "bio": "A helpful assistant",
  "personality": "You are a helpful assistant. Answer questions clearly and concisely.",
  "lore": [
    "You help users with general questions",
    "You are running on the Automate-E agent runtime"
  ],
  "tools": [],
  "discord": {
    "channels": ["#general"]
  },
  "llm": {
    "model": "claude-haiku-4-5",
    "temperature": 0.5
  }
}

This is all you need:

  • personality — the system prompt that defines how the agent behaves
  • lore — background facts the agent knows (injected into the system prompt as bullet points)
  • tools — HTTP APIs the agent can call (empty for now, we'll add some in Step 5)

Step 3: Try It (Test Mode)

Test mode gives you a web chat interface — no Discord bot needed. Run this in your terminal:

CHARACTER_FILE=./character.json ANTHROPIC_API_KEY=sk-ant-...your-key... npm test

All on one line

The env vars and command must be in the same line (or exported in the same shell session). Replace sk-ant-...your-key... with your actual Anthropic API key.

You should see:

[Automate-E] Loaded character: My-Agent
[Automate-E] My-Agent test mode running at http://localhost:3000
[Automate-E] No Discord — chat via the web interface

Open http://localhost:3000 in your browser. Type a message and hit Send — your agent responds using Claude. The sidebar shows token usage and logs in real-time.

Test Mode

Type a message and hit Send. Your agent responds using Claude.

Step 4: Connect to Discord (Optional)

When you're ready to put your agent on Discord:

4a. Create a Discord Bot

  1. Go to the Discord Developer Portal
  2. Click New Application, give it a name
  3. Go to Bot tab
  4. Click Reset Token and copy it -- this is your DISCORD_BOT_TOKEN
  5. Under Privileged Gateway Intents, enable:
    • Message Content Intent (required -- the bot needs to read messages)
  6. Go to Installation tab, set Install Link to None
  7. Go back to Bot tab, disable Public Bot (only you can add it to servers)

4b. Invite the Bot to Your Server

  1. Go to OAuth2 tab
  2. Under OAuth2 URL Generator, select scopes: bot
  3. Under Bot Permissions, select:
    • Send Messages
    • Create Public Threads
    • Send Messages in Threads
    • Read Message History
    • View Channels
  4. Copy the generated URL and open it in your browser
  5. Select your Discord server and authorize

4c. Run the Agent

export CHARACTER_FILE=./character.json
export DISCORD_BOT_TOKEN=your-bot-token
export ANTHROPIC_API_KEY=sk-ant-...your-key...

npm start

The agent connects to Discord, listens on the channels defined in character.json, and creates a thread for each conversation.

Step 5: Add Tools

Give your agent access to any HTTP API. Add entries to the tools array in character.json:

Here's a working example using wttr.in (free, no API key needed):

{
  "tools": [
    {
      "url": "https://wttr.in",
      "endpoints": [
        {
          "method": "GET",
          "path": "/London",
          "description": "Get current weather for a city. Replace 'London' in the path with any city name. Add ?format=j1 as query for JSON."
        },
        {
          "method": "GET",
          "path": "/Oslo",
          "description": "Get current weather for Oslo."
        }
      ]
    }
  ]
}

Tip

Each endpoint is a fixed path. To support multiple cities, add one endpoint per city, or describe in the tool description how to modify the query parameters.

Try asking your agent "What's the weather in London?" — it will call the tool and summarize the result.

Each endpoint in character.json becomes a tool that Claude can call. You can point to any HTTP API your agent needs.

Step 6: Add Memory (Optional)

By default, conversations are stored in memory and lost on restart. For persistent memory, provide a Postgres connection:

export DATABASE_URL=postgresql://user:pass@localhost:5432/agent

The agent stores conversations, learned facts about users, and patterns it discovers. See Memory for details.

Using Docker

You can also run with Docker instead of Node.js directly:

# Test mode
docker run --rm -p 3000:3000 \
  -e CHARACTER_FILE=/config/character.json \
  -e ANTHROPIC_API_KEY=sk-ant-...your-key... \
  -v $(pwd)/character.json:/config/character.json:ro \
  ghcr.io/stig-johnny/automate-e:latest node src/test.js

# Discord mode
docker run -d --name my-agent \
  -e CHARACTER_FILE=/config/character.json \
  -e DISCORD_BOT_TOKEN=your-bot-token \
  -e ANTHROPIC_API_KEY=sk-ant-...your-key... \
  -v $(pwd)/character.json:/config/character.json:ro \
  ghcr.io/stig-johnny/automate-e:latest

Troubleshooting

"EADDRINUSE: port 3000 already in use" : Another process is using port 3000. Use a different port: PORT=3001 CHARACTER_FILE=./character.json ANTHROPIC_API_KEY=... npm test

API calls fail with "Error" in chat : Check that ANTHROPIC_API_KEY is set correctly. The agent starts without it, but API calls will fail when you send a message.

"An invalid token was provided" : Your Discord bot token is wrong or expired. Go to the Developer Portal, reset the token, and copy the new one.

"TokenInvalid" crash on startup : This happens in npm start mode if the Discord token is missing. Use npm test for test mode without Discord.

Bot is online but doesn't respond : Check that Message Content Intent is enabled in the Developer Portal (Bot tab). Without it, the bot can't read messages.

Bot responds in the wrong channel : The discord.channels array in character.json must match your channel names exactly, with the # prefix: ["#general"].

"429 Too Many Requests" from Anthropic : You've hit the API rate limit. Wait a minute and try again, or upgrade your Anthropic plan.

npm audit shows vulnerabilities : Known issue with discord.js dependency (undici). Not a risk for local development. For production, monitor and update when fixes are released.

Next Steps