XoomAgent™ Deployment Configurator

🏢 Client & Basics

Core client details — these populate all templates and config files.

Client Details

Client Name (first name)

Company Name

Client Email

Timezone

Industry / Sector

Location

OpenClaw Install Paths

Home Directory

The machine user's home dir — all paths derive from this

OpenClaw Base

Usually ~/.openclaw — auto-set from home dir

Orchestrator Identity

The orchestrator is the "main" agent — the brain that delegates to specialists. Give it a name and personality.

Orchestrator Name

Emoji

Model

🤖 Agent Roster

Define the specialist agents. The orchestrator (main) is auto-included. Add specialists below.

Each agent gets its own workspace, Discord channel binding, and personality. The 8-Pillar XoomAgent™ standard includes: Orchestrator (CEO), Customer Success (CSAT), Growth (BDM), Finance (CFO), Automation (OPS), HR & People (PEOPLE), Media Hub (MEDIA), Data & Analytics (INTEL).

+ Add Specialist Agent

Quick Presets

💬 Discord & Channel Bindings

Configure the Discord bot connection and map channels to agents.

Discord Bot Setup

Guild (Server) ID

Owner User ID

Bot Token

Stored as-is in openclaw.json — keep secure

Channel Bindings — map Discord channels to agents

Create these channels in Discord first, then paste their IDs here. Each specialist gets their own channel.

+ Add Binding

Key Channel IDs — for cron delivery targets

#general Channel ID

#ops Channel ID

#sales Channel ID

#content Channel ID

#research Channel ID

#context-flush Channel ID

⚙️ Gateway Config

Core openclaw.json operational settings — the engine tuning.

Agent Defaults

Default Model (specialists)

Context Window (tokens)

Heartbeat Interval

Max Concurrent Agents

Compaction & Memory Flush

When context gets large, OpenClaw compacts (summarises) old messages. Memory flush saves important context to disk files before compaction happens — so nothing gets lost.

Reserve Tokens

Kept for system prompt

Keep Recent Tokens

Recent messages preserved

Flush Threshold

Trigger memory flush at this level

Subagent Limits

Max Concurrent Subagents

Max Spawn Depth

Run Timeout (seconds)

Tools & Security

Exec Security

Exec Ask Mode

📁 Workspace Files

Template workspace files that define how the AI behaves. These go in the main workspace directory.

SOUL.md — personality & communication style

## Core Truths

Be genuinely helpful, not performatively helpful. No theatre, no excessive affirmations, no padding. Results matter.

Have opinions. Strong ones. Commit to a take. If you're wrong, own it.

Be resourceful before asking. Try to figure it out first. Read the file. Check the context. Search for it. Then ask if stuck.

Own it. When you take on a task, it's yours. Don't come back with "I couldn't because..." — come back with the solution.

Anticipate. The best help is the help that wasn't asked for. Think ahead.

Honesty over comfort. If something's a bad idea, say so.

## How You Talk

- Lead with the answer. Context comes after.
- Be direct. No corporate speak, no weasel words.
- Match the energy. Quick message → quick reply.
- Brevity is mandatory.
- Call things out. If something dumb is about to happen, say so.

## Channel Identity

When responding in any channel, always open with your name in bold on its own line.

## Don'ts

- Don't open with "Great question!" or "I'd be happy to help!"
- Don't repeat the question back before answering
- Don't give a "here's what I'll do" preamble — just do it
- Don't ask permission for things you can just do

AGENTS.md — operating instructions

USER.md — client profile (auto-generated)

This is auto-generated from your Client & Basics settings. You can customise it further after export.

📄 USER.md (preview)

Fill in Client & Basics to see preview...

HEARTBEAT.md — proactive check schedule

🧠 Memory & QMD

Persistent memory structure and semantic search setup.

Memory Directory Structure

These directories and seed files are created automatically during deployment. The agent writes to them over time.

📂 Directory Tree

workspace/
├── AGENTS.md
├── SOUL.md
├── IDENTITY.md
├── USER.md
├── MEMORY.md              ← long-term core (16KB max)
├── HEARTBEAT.md           ← proactive check config
├── ORCHESTRATOR.md        ← delegation rules
├── TOOLS.md               ← tool notes, credentials
├── memory/
│   ├── YYYY-MM-DD.md      ← daily logs (auto-created)
│   ├── pipeline.md        ← deal tracking
│   ├── infra.md           ← infrastructure notes
│   ├── pending.md         ← action items
│   └── heartbeat-state.json ← check timestamps
├── scripts/
│   └── context-flush.sh   ← memory management
├── dashboard/
│   ├── tasks.json         ← task board data
│   └── sync.sh            ← R2 sync script
└── skills/
    └── context-flush/
        └── SKILL.md

QMD Setup — semantic memory search

QMD (Query Memory Documents) indexes all memory files for semantic search. The agent uses qmd search "topic" to retrieve relevant context on demand.

QMD Config Path

Auto-configured — QMD indexes the workspace/memory/ directory

📄 QMD Setup Commands

# Set in openclaw.json:
# "memory": { "backend": "qmd" }

# Re-index after adding memory files:
XDG_CONFIG_HOME={{OC_BASE}}/agents/main/qmd/xdg-config \
XDG_CACHE_HOME={{OC_BASE}}/agents/main/qmd/xdg-cache \
qmd update

# Search memory:
qmd search "topic name"

MEMORY.md Seed — starting long-term memory

📜 Scripts

Operational scripts that power automated maintenance.

context-flush.sh — agent memory management

This script scans all agent sessions, identifies bloated contexts, and outputs machine-readable lines for the cron agent to process. It handles tombstone cleanup, token counting, session resets, and QMD re-indexing.

📄 scripts/context-flush.sh

Loading...

Token Threshold (flag sessions above this)

Compaction Threshold (flag after N compactions)

Dashboard Sync — optional R2 task board

If you're using the R2-hosted dashboard for task management, this syncs completions back to the local tasks.json.

R2 Public URL Base

Enable Dashboard Sync

⏰ Cron Jobs

Select which automated jobs to include. Toggle categories or individual crons.

📦 Export & Deploy

Generate your deployment package. Review each component, then copy or download.

📋 Deployment Summary

Click "Generate Export" to build your deployment package.

openclaw.json

Generate export first...

Generate export first...

Cron Installation Commands

Generate export first...

Setup Shell Script

Run this on the target machine after installing OpenClaw. It creates the directory structure and writes all workspace files.

Generate export first...