ironclaw

Configuration and infrastructure for IronClaw, Ironhack's internal AI assistant running on a self-hosted OpenClaw server.

---

Quick reference

Item	Value
SSH alias	openclaw-server
Server hostname	openclaw-ironhack
OS	Ubuntu 24.04.4 LTS
OpenClaw version	2026.5.4
Gateway port	18789
Service	openclaw-gateway.service (user systemd, openclaw user)

Useful commands:

systemctl --user status openclaw-gateway
systemctl --user restart openclaw-gateway
journalctl --user -u openclaw-gateway -n 50
openclaw sessions --all-agents --active 60

---

Agents

Four agents, all running in the same gateway process:

Agent	Channel	Purpose
ironclaw	DMs + #ironclaw-watch	Router, leadership ops, competitive intelligence
ironclaw-seo	#ironclaw-seo	SEO analyst
ironclaw-edu	#ironclaw-edu	Education content QA
ironclaw-jobs	#ironclaw-jobs	German job market research for caseworker reports

• #ironclaw-watch routes to ironclaw with a channel-level system prompt that primes it for competitive analysis. Also receives the Monday 07:00 Rome cron digest.
• DMs go to ironclaw via pairing (no binding needed). Command owner: Rudy (slack:U02MV9VPGV6).

Workspace files live in server/workspace-ironclaw-*/ in this repo. Deploy with ./scripts/deploy-workspaces.sh.

---

Cron jobs

Job	Agent	Schedule	Delivery
Weekly Competitor Watch	ironclaw	Mon 07:00 Rome	#ironclaw-watch
Weekly SEO Report	ironclaw-seo	Mon 08:00 Rome	#ironclaw-seo
Course Content Review	ironclaw-edu	Daily 06:00 Rome	#ironclaw-edu
Weekly Job Scrape	ironclaw-jobs	Mon 09:00 Rome	#ironclaw-jobs
Weekly Caseworker Report	ironclaw-jobs	Mon 10:00 Rome	#ironclaw-jobs

---

Server layout

/home/openclaw/.openclaw/
  openclaw.json               # Main config (agents, bindings, channels, plugins)
  gateway.systemd.env         # All secrets (loaded by systemd)
  .env                        # Legacy secrets file (AWS keys; keep in sync)
  agents/
    ironclaw/
      agent/                  # Auth profiles, model overrides
      sessions/               # Session history
    ironclaw-seo/
    ironclaw-edu/
  workspace/                  # ironclaw agent workspace
    SOUL.md                   # Core persona
    AGENTS.md                 # Session startup rules
    TOOLS.md                  # Environment-specific notes (Playwright, S3)
    MEMORY.md                 # Long-term curated memory
    USER.md                   # User profile
    COMPETITORS.md            # Competitor tracking list
    skills/
      competitor-watch/       # Competitor scan + report skill
      course-reviewer/        # Course QA skill
    competitor-snapshots/     # Weekly JSON snapshots (delta source)
    memory/                   # Daily session logs
  workspace-ironclaw-seo/     # ironclaw-seo workspace (deploy from repo)
  workspace-ironclaw-edu/     # ironclaw-edu workspace (deploy from repo)
  workspace-ironclaw-jobs/    # ironclaw-jobs workspace (deploy from repo)
  cron/
    jobs.json                 # Cron job definitions
  credentials/                # Channel credentials (pairing, allowlists)

---

Stack

Component	Detail
Model	ZAI GLM-5.1 (thinking=medium) + GLM-5-turbo / GLM-4.7 fallbacks
Memory	OpenAI embeddings (vector search)
Search	Tavily (primary) + DuckDuckGo (fallback)
Web fetch	Firecrawl (key pending)
Browser	Playwright-core + Chromium 1217 (/home/openclaw/.cache/ms-playwright/)
S3	ih-ironclaw bucket, eu-west-1
Slack	Socket Mode, App ID A0B19LV97DM
Telegram	Disabled

---

Secrets (not committed)

All secrets in /home/openclaw/.openclaw/gateway.systemd.env:

Variable	Status
OPENCLAW_GATEWAY_TOKEN	Set
OPENAI_API_KEY	Set (used for memory embeddings)
SLACK_BOT_TOKEN / SLACK_APP_TOKEN	Set (App A0B19LV97DM)
AWS_ACCESS_KEY_ID / AWS_SECRET_ACCESS_KEY	Set
AWS_DEFAULT_REGION	eu-west-1
TAVILY_API_KEY	Set
FIRECRAWL_API_KEY	PLACEHOLDER — get from firecrawl.dev
COMPOSIO_API_KEY	PLACEHOLDER — needed for GSC integration
GOG_KEYRING_PASSWORD	Set

---

Pending

• Firecrawl API key — replace PLACEHOLDER in gateway.systemd.env, restart gateway
• Education agent bootstrap — populate review-queue.json before enabling daily cron
• Composio + GSC — needed for SEO agent's Search Console access
• S3 bucket policy — add jobs/* to public-read policy on ih-ironclaw bucket (via AWS console; IAM user lacks s3:GetBucketPolicy). Current policy covers seo/*, edu/*, ironclaw/*, watch/*. Add: arn:aws:s3:::ih-ironclaw/jobs/*