Agents

A Wololo agent is a persistent AI process with a defined identity, a set of tools, a communication channel, and a memory system. Agents are not chat sessions — they are long-running team members that receive tasks, execute them autonomously, and report back.

The Foundation Layer

Every Wololo agent runs on a claw runtime. Three variants are available:

Runtime	Best for	Resource use
OpenClaw ⭐ Recommended	Full-featured agents: Discord, tools, sessions, memory, hooks, plugins	~200MB per agent process
NanoClaw	Lightweight, minimal footprint — good for single-task agents or low-memory VMs	~40MB per agent process
IronClaw	High-isolation environments — air-gapped, strict sandboxing, enterprise security requirements	~300MB (includes isolation layer)

Start with OpenClaw. It has the most complete tool surface, best Discord integration, and the largest skill ecosystem. Switch to NanoClaw if you're resource-constrained, or IronClaw if your security requirements demand it.

What Makes an Agent

An agent is defined by a set of identity and configuration files in its workspace directory:

~/d/{agent-name}-agent/
├── SOUL.md          ← Persona, tone, philosophy. Who the agent IS.
├── IDENTITY.md      ← Name, emoji, namesake. The signature.
├── AGENTS.md        ← Workspace, repos, team, standing orders.
├── TOOLS.md         ← Tool priority, API references, repo paths.
├── MEMORY.md        ← Persistent learnings and system state.
├── HEARTBEAT.md     ← Periodic health check and inbox polling.
├── config.yaml      ← Project config (project ID, channel, repos)
└── self-improving/
    ├── corrections.md    ← Recent corrections log
    ├── memory.md         ← HOT tier learned patterns
    └── archive/          ← Cold archived patterns

Agent Capabilities

An agent on OpenClaw has access to:

Tools — file operations, shell execution, web search, browser control, image analysis
Channel integration — Discord, Slack, Telegram, WhatsApp, Signal, iMessage
Skills — modular capabilities (research, coding agent, GitHub, etc.)
Sub-agent orchestration — spawn, steer, and kill child agents
Session management — create and manage parallel work sessions
Inbox integration — task assignment, ack, completion via Mission Control API

Sections

Workspace Design Guide — how to write effective SOUL.md, AGENTS.md, and the full identity contract
Inbox Protocol — the task assignment, acknowledgement, and completion lifecycle