T420 OpenClaw Autonomous Capital Raise Agent

Autonomous agent deployment for T420's capital raise campaign. Two tracks running in parallel:

• Track A: $5,000 token sales via KOL partnerships, X/Twitter, Telegram, launchpads

• Track B: $1.5M seed round via VC warming + cold email outreach

---

Architecture

ZANE (Founder)
├── Codes FleetOS, Robo-Hub, Robo-Dapp
├── Reviews daily brief (15 min/morning)
├── Takes VC calls when booked
└── Course-corrects when needed

OPENCLAW AGENT (Runs 24/7 on AWS EC2)
├── t420-kol-scout     → Find KOLs, send personalized DMs
├── t420-vc-outreach   → Warm VCs on Twitter, send cold emails
├── t420-x-engine      → Post threads, engage in conversations
├── t420-launchpad-hunter → Apply to launchpads + ICO directories
├── t420-telegram-presence → Community engagement in external Telegram groups
├── t420-community-bot → Our @t420_community group: announcements + Q&A
└── t420-daily-brief   → Morning intelligence report

---

Files

openclaw/
├── SOUL.md                              # Agent identity, brand, guardrails
├── setup.sh                             # AWS EC2 setup script
├── .env.example                         # API key template
├── README.md                            # This file
├── skills/
│   ├── t420-kol-scout/SKILL.md         # KOL discovery + DM
│   ├── t420-vc-outreach/SKILL.md       # VC email campaign
│   ├── t420-x-engine/SKILL.md          # X/Twitter content + engagement
│   ├── t420-launchpad-hunter/SKILL.md  # Launchpad applications
│   ├── t420-telegram-presence/SKILL.md # External Telegram community engagement
│   ├── t420-community-bot/SKILL.md     # Our own @t420_community group bot
│   └── t420-daily-brief/SKILL.md       # Morning report
└── data/
    ├── vc-targets.json                  # 14 VC firms (from existing CSV)
    ├── thread-queue.json                # 4 pre-approved threads
    └── kol-tracker.json                 # KOL outreach log (starts empty)

---

Deployment Guide

Prerequisites

• AWS account (free tier eligible)

• ChatGPT Plus ($20/mo) or Pro ($200/mo) subscription — powers all LLM calls via Codex OAuth

• API keys for: Brave Search, X/Twitter API v2, SMTP email

• Optional: Grok API, Exa, Firecrawl, Apify, Telegram bot token, OpenAI API key (for embeddings only)

Step 1: Launch EC2 Instance

1. Go to AWS Console > EC2 > Launch Instance

2. Choose Ubuntu 22.04 LTS (free tier eligible)

3. Instance type: t2.micro (free tier)

4. Storage: 20 GB gp3

5. Security group: Allow SSH (port 22) from your IP only

6. Create/select key pair, launch instance

7. SSH in: ssh -i your-key.pem ubuntu@<public-ip>

Step 2: Run Setup Script

# Upload files to EC2
scp -i your-key.pem -r openclaw/ ubuntu@<public-ip>:~/

# SSH in and run setup
ssh -i your-key.pem ubuntu@<public-ip>
cd ~/openclaw
chmod +x setup.sh
./setup.sh

Step 3: Authenticate with ChatGPT (Codex OAuth)

# Authenticate using your ChatGPT subscription (opens browser)
openclaw models auth login --provider openai-codex

# Set the model
openclaw models set openai-codex/gpt-5.3-codex

# Verify it worked
openclaw models status

If gpt-5.3-codex gives errors, fall back to openai-codex/gpt-5.2-codex.

Step 4: Configure Remaining API Keys

cp .env.example ~/openclaw/.env
nano ~/openclaw/.env
# Fill in non-LLM API keys (X/Twitter, Brave, SMTP, Telegram)

Minimum required keys:

• X_API_KEY + related tokens — X/Twitter posting and DMs

• BRAVE_SEARCH_API_KEY — Web search for research

• SMTP_* — VC cold email sending

• T420_BOT_TOKEN — Telegram community bot

Step 5: Security Audit Skills

Before starting, review every ClawHub skill's source code:

openclaw skill inspect brave-search
openclaw skill inspect chirp
openclaw skill inspect smtp-send
# ... review each one

Step 6: Test Each Skill

# Test individually before going live
openclaw skill test t420-kol-scout      # Should find KOLs, NOT send DMs
openclaw skill test t420-x-engine       # Should draft content, NOT post
openclaw skill test t420-vc-outreach    # Should research VCs, NOT email
openclaw skill test t420-daily-brief    # Should generate sample brief

Step 7: Activate (Go Live)

# Start everything with PM2
cd ~/openclaw
pm2 start ecosystem.config.js
pm2 save                              # auto-start on reboot

# Monitor logs
pm2 logs t420-openclaw-agent
pm2 logs t420-telegram-bot

---

Activation Sequence

Start skills in this order, with 24-hour observation between each:

Day	Activate	Monitor For
1	t420-daily-brief	Brief generates correctly
1	t420-x-engine (posting only)	Content posts correctly, no errors
2	t420-x-engine (engagement)	Replies are on-brand, not flagged
2	t420-telegram-presence (listen mode)	Joins groups, no issues
3	t420-kol-scout	DMs send correctly, spacing enforced
5	t420-telegram-presence (contribute mode)	Messages are natural
5	t420-vc-outreach (warming only)	Twitter engagement is appropriate
7	t420-launchpad-hunter	Applications submit correctly
10	t420-vc-outreach (emails)	Emails deliver, not marked as spam

---

Daily Operations

Morning Routine (Zane — 15 min)

1. Read briefs/YYYY-MM-DD-brief.md

2. Check URGENT section — action anything flagged

3. Review metrics — are we on track?

4. Approve/reject any pending actions

5. Done. Go code.

Weekly Review (Zane — 30 min, Sundays)

1. Read weekly digest section in brief

2. Review what's working vs. not

3. Add new threads to thread-queue.json if queue is low

4. Adjust strategy based on data

5. Check agent health and error logs

---

Kill Switch

Stop Everything Immediately

pm2 stop all

Stop Specific Process

pm2 stop t420-openclaw-agent    # Stop agent, keep bot running
pm2 stop t420-telegram-bot      # Stop bot, keep agent running

Stop Specific Skill

openclaw skill disable t420-kol-scout
openclaw skill disable t420-x-engine
# etc.

Emergency: Account Compromised

# Stop everything
pm2 stop all

# Revoke API keys immediately
# 1. X/Twitter: https://developer.twitter.com → revoke app
# 2. Rotate SMTP password
# 3. Rotate OpenAI API key
# 4. Rotate all keys in .env

---

LLM Configuration (Codex OAuth)

The agent uses ChatGPT subscription via Codex OAuth — no per-token API billing.

Provider	Model	Auth
OpenAI (Codex OAuth)	gpt-5.3-codex	ChatGPT Plus/Pro subscription

Setup:

openclaw models auth login --provider openai-codex
openclaw models set openai-codex/gpt-5.3-codex
openclaw models status   # verify

Rate Limits (ChatGPT Plus — $20/month):

• ~45–225 messages per 5-hour rolling window

• ~10–60 cloud tasks per 5-hour window

• Session-based execution keeps usage well within limits

Fallback: If gpt-5.3-codex is unavailable, use openai-codex/gpt-5.2-codex.

Embeddings: Codex OAuth does not cover text embeddings. If needed, add OPENAI_EMBEDDINGS_API_KEY in .env for pay-as-you-go embeddings only.

---

Rate Limits Summary

Channel	Limit	Spacing
X DMs	20/day (hard cap 30)	30 min apart
X posts	3/day	2 hours apart
X replies	20/day	15 min apart
X likes	50/day	—
Cold emails	5/day	2 hours apart
Telegram messages	15/group/day	—
Telegram groups/hour	5	—
Launchpad apps	2/day	—

---

Guardrails (Non-Negotiable)

Full guardrails are in SOUL.md. Key points:

The agent CANNOT:

• Promise returns or guarantees

• Lie about traction or status

• Access wallet keys or treasury

• Change token price/terms

• Respond to legal/regulatory questions

• Be aggressive, spammy, or desperate

The agent MUST:

• Include NFA/DYOR on investment content

• Escalate VC positive responses to Zane

• Escalate legal/regulatory inquiries to Zane

• Respect all rate limits

• Report honestly in daily briefs

---

KPIs

Track A: $5,000 Token Sales (3-week targets)

Metric	Week 1	Week 2	Week 3
KOL DMs sent	70	140	200+
KOL response rate	8%	12%	15%
KOL deals closed	2-3	5-7	8-10
X followers gained	+100	+300	+500
Revenue	$500	$2,000	$5,000

Track B: $1.5M Seed Round

Metric	Week 1	Week 2	Week 3
VC warming engagements	15	40	80+
Cold emails sent	4	10	20
Responses	0-1	2-4	5-8
Calls booked	0	1-2	3-5

---

Troubleshooting

Issue	Fix
Agent not starting	Check .env keys, run openclaw doctor
DMs not sending	Verify X API tokens, check rate limits
Emails bouncing	Verify SMTP config, check SPF/DKIM
Brief not generating	Check cron: crontab -l, check pm2 logs
Skill errors	openclaw skill test <name>, check skill logs
Account restricted	STOP agent (pm2 stop all), wait 24h, contact platform support
EC2 out of memory	pm2 restart all, consider t3.micro upgrade
PM2 processes not restarting	pm2 startup, pm2 save