Automate-E Runtime¶
Automate-E is a Kubernetes-native AI agent runtime for Discord. It turns a character.json config file into a running Discord bot backed by Claude, with persistent memory, tool calling, and cost tracking.
- Repository: Stig-Johnny/automate-e (public, MIT license)
- Documentation: stig-johnny.github.io/automate-e
- Deployment: Helm chart in
charts/automate-e/, deployed via ArgoCD
Key Features¶
| Feature | Description |
|---|---|
| Character-driven | Define personality, tools, and behavior in a single JSON file |
| Claude-powered | Agent loop uses Anthropic Claude API with tool use |
| Postgres memory | Conversations, user facts, and merchant patterns persist across restarts |
| In-memory fallback | Works without Postgres for local development |
| Tool calling | Agents call HTTP APIs defined in character.json |
| Cost tracking | Per-model token usage and cost calculation |
| Live dashboard | Real-time WebSocket dashboard for monitoring agent activity |
| Helm chart | Deploy single-process or gateway+worker mode via Helm |
| Split mode | Gateway + Redis Streams + workers for horizontal scaling |
| Webhook receiver | Receive real-time events from GitHub, CI systems, etc. |

How It Works¶
sequenceDiagram
participant U as Discord User
participant G as Discord Gateway
participant A as Agent Loop
participant C as Claude API
participant T as Tool APIs
participant M as Memory (Postgres)
U->>G: Message in #example-e
G->>A: messageCreate event
A->>M: Load conversation history + user facts
A->>C: System prompt + history + tools
C->>A: tool_use: get_quotes_random
A->>T: GET /quotes/random
T->>A: {text: "Talk is cheap...", author: "Linus Torvalds"}
A->>C: Tool result
C->>A: Text response
A->>M: Save messages
A->>G: Reply in thread
G->>U: "Here's a quote from Linus Torvalds: ..."
Get Started¶
- Quick Start -- run an agent locally in 5 minutes (test mode, no Discord needed)
- Discord Bot Setup -- create a bot and connect to Discord
- Configuration -- full
character.jsonreference
Learn More¶
- Architecture -- how the runtime works internally
- Deployment -- deploy to Kubernetes with Helm
- Dashboard -- real-time monitoring UI
- Memory -- persistent conversations and facts