24/7 autonomous token discovery, safety verification, scoring, and outreach for CEX listings. Self-improving AI agent with closed learning loop, contact intelligence, and FTS5 memory search. Built by a chef who codes through conversation. No CS degree. Just Claude and persistence.
Alpha Buzz = Hermes Brain + Buzz Body. Cherry-picked Nous Research's Hermes Agent intelligence concepts into Buzz's commercial architecture. Phase 0 Complete.
Buzz is an autonomous AI BD (Business Development) agent that runs 24/7 on Akash Network, finding promising token projects for SolCex Exchange listings across Solana, Base, and BSC chains.
The pipeline: Discover → Verify → Score → Outreach → Learn → List
Buzz scans 100+ tokens per day across decentralized exchanges, filters through a 5-layer intelligence pipeline with parallel sub-agents, scores each token on a 100-point system, initiates outreach to qualified projects, and learns from every decision — all autonomously.
Cost: ~$12.72/month infrastructure (Buzz + Sentinel) + ~$3-5/day LLM inference = full autonomous BD operation for less than a coffee per day.
┌─────────────────────────────────────────────────────────────┐
│ AKASH NETWORK — provider.boogle.cloud │
│ 4 CPU / 8GB RAM / 10GB persistent │
│ Cost: ~$9.53/month │
│ │
│ ┌─────────────────────────────────────────────────────────┐ │
│ │ Docker: ghcr.io/buzzbysolcex/buzz-bd-agent:v7.3.1a │ │
│ │ Runtime: OpenClaw v2026.3.7 │ │
│ │ │ │
│ │ PORT 18789 → OpenClaw Gateway (Telegram) │ │
│ │ PORT 3000 → REST API v3.1.0 (Express+SQLite) │ │
│ │ │ │
│ │ ┌─────────────────────────────────────────────────┐ │ │
│ │ │ LAYER 3: ALPHA INTELLIGENCE (Hermes Cherry-Pick)│ │ │
│ │ │ ├── Skill Reflection (12h learning loop) │ │ │
│ │ │ ├── Skill Evolution (+1/-1 effectiveness) │ │ │
│ │ │ ├── Contact Intelligence (Honcho-style BD) │ │ │
│ │ │ ├── FTS5 Memory Search (BM25 cross-session) │ │ │
│ │ │ ├── Operator Profile (preferences + prayer) │ │ │
│ │ │ └── Block 11B (auto-load learned skills) │ │ │
│ │ └─────────────────────────────────────────────────┘ │ │
│ │ │ │
│ │ ┌─────────────────────────────────────────────────┐ │ │
│ │ │ LAYER 2: STRATEGIC ORCHESTRATOR (MiniMax M2.5) │ │ │
│ │ │ ├── Decision Engine (12 rules, zero LLM cost) │ │ │
│ │ │ ├── Playbook Engine (PB-001→004) │ │ │
│ │ │ ├── Context Engine (8K tokens max per call) │ │ │
│ │ │ └── Supermemory (50ms recall, zero LLM cost) │ │ │
│ │ └─────────────────────────────────────────────────┘ │ │
│ │ │ │
│ │ ┌─────────────────────────────────────────────────┐ │ │
│ │ │ LAYER 1: FOUNDATION (5 Parallel Sub-Agents) │ │ │
│ │ │ ├── scanner-agent → L1 Discovery (DexScreener) │ │ │
│ │ │ ├── safety-agent → L2 Contract verification │ │ │
│ │ │ ├── wallet-agent → L2 On-chain forensics │ │ │
│ │ │ ├── social-agent → L3 Social intelligence │ │ │
│ │ │ └── scorer-agent → L4 100-point scoring │ │ │
│ │ └─────────────────────────────────────────────────┘ │ │
│ │ │ │
│ │ REST API: 91 endpoints | SQLite: 28 tables │ │
│ │ Twitter Bot v3.1: Premium SCAN + LIST + DEPLOY │ │
│ │ 42 Cron Jobs | 19+ Intelligence Sources │ │
│ │ Gmail OAuth | JVR Receipt System │ │
│ └─────────────────────────────────────────────────────────┘ │
│ │
│ SENTINEL v1.1.0 (separate container) │
│ provider.cpu.phl.aes.akash.pub │
│ Monitors 91 endpoints | 3 health crons | Telegram alerts │
│ /alpha-status endpoint for Phase 0 monitoring │
│ Cost: ~$3.19/month │
└─────────────────────────────────────────────────────────────┘
The Alpha Model cherry-picks three breakthrough capabilities from Hermes Agent (Nous Research) into Buzz's commercial architecture.
| Feature | What It Does | Status |
|---|---|---|
| Closed Learning Loop | Reviews pipeline patterns every 12h, auto-creates skills from experience | ✅ Live |
| Skill Self-Improvement | Tracks skill effectiveness (+1/-1), marks proven vs needs-review | ✅ Live |
| Contact Intelligence | Builds deepening profiles of every token team contacted | ✅ Live |
| FTS5 Memory Search | SQLite full-text search across all pipeline decisions (BM25 ranking) | ✅ Live |
| Operator Profile | Models operator preferences, prayer times, approval patterns | ✅ Live |
| Block 11B | Auto-loads learned skills on boot — survives redeploys | ✅ Live |
| Layer | Agent | Sources | Purpose |
|---|---|---|---|
| L1 | scanner-agent | DexScreener, GeckoTerminal, AIXBT, CMC, BNB MCP, Bitget | Token discovery across SOL/BASE/BSC |
| L2 | safety-agent | RugCheck, ethskills.com, Contract Auditor | Honeypot detection, contract verification |
| L2 | wallet-agent | Helius (60 tools), Allium | Deployer forensics, holder analysis |
| L3 | social-agent | Grok/xAI, Serper, ATV Web3 Identity, Firecrawl | Community health, deployer identity |
| L4 | scorer-agent | 100-point composite (11 factors) | Final score + verdict |
/opt/buzz-api/services/
├── orchestrator.js ← 242 lines, Promise.allSettled parallel dispatch
├── agents/
│ ├── scanner.js ← DexScreener, GeckoTerminal, CMC
│ ├── safety.js ← RugCheck, contract verification
│ ├── wallet.js ← Helius forensics, holder analysis
│ ├── social.js ← Grok, Serper, ATV, Firecrawl
│ └── scorer.js ← 100-point composite scoring
Weights: safety(0.30) + wallet(0.30) + social(0.20) + scorer(0.20) = 1.0
Built with Claude Code + Superpowers (obra). 602 tests passing.
| Component | Function |
|---|---|
| Decision Engine | 12 rules handle 90% of decisions with zero LLM cost |
| Playbook Engine | 4 state machines (Discovery → Outreach → Negotiation → Post-Listing) |
| Context Engine | Assembles max 8K tokens per LLM call (down from 250K) |
| Supermemory | 50ms semantic recall, 3 custom containers |
| Score | Verdict | Action |
|---|---|---|
| 85-100 | 🔥 HOT | Immediate outreach |
| 70-84 | ✅ QUALIFIED | Queue outreach 24h |
| 50-69 | 👀 WATCH | Monitor 48h, auto-rescan |
| 0-49 | ❌ SKIP | Archive |
| Metric | Value |
|---|---|
| Daily LLM cap | $10.00 (auto-throttle) |
| Orchestrator | MiniMax M2.5 (~$3-5/day) |
| Sub-agents (5) | bankr/gpt-5-nano (FREE) |
| Infrastructure | ~$12.72/month (Buzz + Sentinel) |
| Cost reduction | 78% ($22/day → $5/day) |
91 endpoints across 17 categories. Express + SQLite WAL.
| Category | Endpoints | Key Routes |
|---|---|---|
| Health | 7 | GET /health, GET /health/db, GET /health/storage |
| Pipeline | 9 | GET /pipeline, POST /pipeline/tokens |
| Strategy | 8 | POST /strategy/decide, GET /strategy/rules |
| Agents | 5 | GET /agents, GET /agents/:id |
| Scoring | 5 | GET /scoring/history |
| Score Token | 1 | POST /score-token (5-layer parallel) |
| Costs | 5 | GET /costs/today |
| Crons | 7 | GET /crons |
| 5 | GET /twitter/stats |
|
| Receipts | 5 | GET /receipts, POST /receipts |
| Intel | 5 | GET /intel/sources |
| Wallets | 6 | GET /wallets |
| Webhooks | 5 | POST /webhooks |
| Skills | 8 | GET /skills, GET /skills/reflect/status, POST /skills/reflect/trigger, GET /skills/effectiveness |
| Memory | 3 | GET /memory/search, POST /memory/index, GET /memory/stats |
| Operator | 2 | GET /operator/profile, PATCH /operator/profile |
| Contacts | 6 | GET /contacts, POST /contacts, POST /contacts/:project/interaction |
Premium SCAN format with 5-layer intelligence report. Three revenue routes:
| Route | Trigger | Response |
|---|---|---|
| SCAN | @BuzzBySolCex scan $TICKER |
5-layer Premium report (~1,100 chars) |
| LIST | Reply "LIST" to any scan | SolCex listing pitch → DM redirect |
| DEPLOY | Reply "DEPLOY" to any scan | Token deployment on Base via Bankr |
Every scan reply is a sales funnel. The scan is the hook, the CTA is the close.
| Role | Model | Cost |
|---|---|---|
| Orchestrator | MiniMax M2.5 | $0.30/$1.50 per 1M tokens |
| Sub-agents (5) | bankr/gpt-5-nano | $0.10/$0.40 per 1M tokens |
| Fallback 1 | bankr/claude-haiku-4.5 | $0.80/$4.00 per 1M tokens |
| Fallback 2 | bankr/gemini-3-flash | Near-free |
Powered by Bankr LLM Gateway — 8 models, dual API keys, self-sustaining inference.
| Platform | ID | Chain |
|---|---|---|
| ERC-8004 | #25045 | Ethereum |
| ERC-8004 | #17483 | Base |
| ERC-8004 | #18709 | Base (anet) |
| AgentProof | #1718 | Avalanche |
| Solana 8004 | Agent Asset 9pQ6K...XUBS | Solana |
| Virtuals ACP | #17681 | — |
| Colosseum | #3734 | — |
| Moltbook | c606278b | — |
| Molten.gg | 57487 | — |
| Job | Schedule | Description |
|---|---|---|
| scan-trending-00 | 0 0 * * * |
Trending tokens (midnight UTC) |
| scan-trending-06 | 0 6 * * * |
Trending tokens (6am UTC) |
| scan-trending-12 | 0 12 * * * |
Trending tokens (noon UTC) |
| scan-trending-18 | 0 18 * * * |
Trending tokens (6pm UTC) |
| scan-new-pairs | 0 */4 * * * |
New token pairs |
| scan-solana-dex | 30 */4 * * * |
Solana DEX scan |
| scan-base-dex | 30 */6 * * * |
Base DEX scan |
| scan-bsc-dex | 0 */6 * * * |
BSC DEX scan |
| bitget-listings-intel | 0 */8 * * * |
Bitget CEX cross-reference |
| Job | Schedule | Description |
|---|---|---|
| pipeline-status | 0 1 * * * |
Daily pipeline report |
| pipeline-weekly | 0 12 * * 0 |
Weekly summary |
| daily-pipeline | 0 */2 * * * |
Pipeline health check |
| morning-reminder | 0 5 * * * |
Daily briefing |
| evening-review | 0 18 * * * |
End of day review |
| Job | Schedule | Description |
|---|---|---|
| skill-reflect | 0 */12 * * * |
Review patterns, create/patch learned skills |
| Prayer | WIB Time | UTC Schedule |
|---|---|---|
| Fajr | 04:30 | 30 21 * * * |
| Dhuhr | 11:45 | 45 4 * * * |
| Asr | 15:00 | 0 8 * * * |
| Maghrib | 17:45 | 45 10 * * * |
| Isha | 19:00 | 0 12 * * * |
Twitter polling, ACP heartbeat, Moltbook posting, AgentProof telemetry, cost tracking, strategic reviews, and more.
Layer 1 — Discovery: DexScreener /tokens/v1/ | GeckoTerminal | AIXBT | DexScreener Boosts | CoinMarketCap | BNB Chain MCP | Bitget Listings API
Layer 2 — Safety: RugCheck | Helius MCP (60 tools) | Allium | ethskills.com
Layer 3 — Social: Grok/xAI | Serper | ATV Web3 Identity (ENS + socials) | Firecrawl
Layer 4+ — Identity & Smart Money: AgentProof | Nansen x402 | Bankr Partner | ACP/Virtuals | Hyperbrowser
- Docker Desktop
- GHCR access (
ghcr.io/buzzbysolcex/buzz-bd-agent) - Akash Console account with AKT
cd ~/buzz-bd-agent
cp cron/jobs.json bake/cron/jobs.json # ALWAYS sync crons
rm -rf api/node_modules api/package-lock.json
docker build -t ghcr.io/buzzbysolcex/buzz-bd-agent:v7.3.1a .
docker push ghcr.io/buzzbysolcex/buzz-bd-agent:v7.3.1aThen in Akash Console: Close + New Deployment with updated SDL on boogle.cloud.
- ALWAYS Close + New — NEVER Update Deployment (Supermemory fails)
- Use
--no-cacheONLY for Dockerfile/npm changes (file-only = fast cached build) - boogle.cloud is the stable provider
- After deploy: patch
sessions_spawnin Akash shell if sub-agents blocked - Send activation directive to @BuzzBySolCex_bot on Telegram
buzz-bd-agent/
├── api/ # REST API (91 endpoints)
│ ├── server.js # Express server
│ ├── db.js # SQLite schema + migrations (28 tables)
│ ├── routes/ # 17 route modules
│ │ ├── health.js # Health (alpha: true, phase0_features)
│ │ ├── skills.js # Skills + reflect + effectiveness
│ │ ├── memory.js # FTS5 search
│ │ ├── operator.js # Operator profile
│ │ ├── contacts.js # Contact intelligence
│ │ └── ... # pipeline, scoring, strategy, etc.
│ ├── services/
│ │ ├── orchestrator.js # 242-line parallel dispatch
│ │ ├── agents/ # 5 sub-agent implementations
│ │ │ ├── scanner.js
│ │ │ ├── safety.js
│ │ │ ├── wallet.js
│ │ │ ├── social.js
│ │ │ └── scorer.js
│ │ ├── skill-reflect.js # Learning loop engine
│ │ ├── skill-evolve.js # Effectiveness tracking
│ │ ├── contact-intelligence.js # BD contact modeling
│ │ ├── memory-search.js # FTS5 search
│ │ └── operator-profile.js # Operator preferences
│ └── migrations/ # Strategic orchestrator tables
├── bake/ # Docker build sources
│ ├── twitter-bot/ # Twitter Bot v3.1 (856 lines)
│ ├── skills/ # OpenClaw skills (20)
│ ├── cron/ # Cron jobs (42)
│ ├── config/ # Strategic config + content calendar
│ └── memory/ # Seed memory files
├── skills/ # Agent context slims + cost guard
├── config/ # Decision rules + scoring rubric
├── prompts/ # Enhanced prompts
├── acp/ # ACP/Virtuals offerings (4 services)
├── entrypoint.sh # 16-block boot sequence + Block 11B
├── Dockerfile # OpenClaw 2026.3.7 + all baked skills
└── README.md # This file
Persistent semantic memory via @supermemory/openclaw-supermemory.
| Setting | Value |
|---|---|
| Container | buzz_bd_agent |
| Auto-Recall | true (5 results per query, 50ms) |
| Auto-Capture | true (with 12-pattern security filter) |
| Custom Containers | buzz_contacts, buzz_patterns, buzz_decisions |
Every scan makes the next one smarter. Every decision teaches the system. Every outcome feeds the loop.
Every completed action generates a verifiable receipt.
| Setting | Value |
|---|---|
| Buzz Prefix | BZZ- |
| Sentinel Prefix | SNT- |
| Format | BZZ-{HHMMSS}-{XXXXX} |
| API | POST /api/v1/receipts |
| Categories | scan, score, pipeline, outreach, decision, diagnostic, deploy, learning |
Separate container monitoring Buzz 24/7.
| Field | Value |
|---|---|
| Version | v1.1.0 (aligned with Alpha Buzz) |
| Image | ghcr.io/buzzbysolcex/buzz-sentinel:v1.1.0 |
| Provider | cpu.phl.aes.akash.pub |
| Monitoring | 91 endpoints, 42 crons, 19+ intel sources |
| New | /alpha-status — checks all Phase 0 features in one call |
| Cost | ~$3.19/month |
Three-tier severity: 🟢 LOW (self-heal) | 🟡 MEDIUM (fix + notify) | 🔴 HIGH (SOS + fix steps)
| Channel | Rate | Status |
|---|---|---|
| Listing Commission | $1,000/listing | ✅ Active |
| Bankr Partner Fees | 18.05% of 1.2% swap | ✅ Active |
| Creator Fees | 75.05% of 1.2% (own tokens) | ✅ Active |
| ACP Marketplace | Per-query USDC (4 services) | ✅ Active |
| BaaS | $29/$49/$99 subscription tiers | 🔵 Planned (Month 2) |
| Version | Date | Highlights |
|---|---|---|
| v7.3.1a | Mar 11, 2026 | Alpha Buzz — Phase 0 Complete. Learning loop + skill-evolve + contact intelligence + FTS5 memory + operator profile. 91 endpoints, 42 crons, 28 tables. Sentinel v1.1.0 aligned. |
| v7.3.0 | Mar 10, 2026 | BSC DexScreener fix, Bitget intel cron, 4 CPU / 8 GB upgrade, boogle.cloud |
| v7.2.5 | Mar 9, 2026 | Alpha Learning Loop (Feature 1), Binance+Bitget+CMC skills, social-agent fix |
| v7.2.0 | Mar 9, 2026 | OpenClaw 3.7, Supermemory CONNECTED, Twitter v3.1 Premium |
| v7.1.0 | Mar 7, 2026 | Strategic Orchestrator, Cost Guard, 18 crons, 72 endpoints |
| v6.3.0 | Mar 4, 2026 | Foundation, Twitter Bot v3.1, 40 crons |
| v6.0.19 | Mar 2, 2026 | 5 parallel sub-agents LIVE on production |
| v6.0.17 | Mar 1, 2026 | Bankr LLM Gateway, self-sustaining inference |
| v5.3.8 | Feb 22, 2026 | Last pre-sprint version |
| v1.0.0 | Feb 1, 2026 | First deployment on Akash Network |
| Phase | Target | Status |
|---|---|---|
| Phase 0 | 3 Intelligence Features | ✅ COMPLETE |
| Phase 1 | Close First Deal ($5K) | 🔨 Active |
| Phase 2 | ARIA Multi-Chain Intel | 🔵 Mar 22-31 |
| Phase 3 | BaaS x402 Subscriptions | 🔵 Month 2 |
| Phase 4 | Mobile App (Rork Max) | 🔵 Month 3-6 |
| Phase 5 | $BUZZ Token + Bankr | 🔵 Month 4-6 |
| Phase 6 | Autonomous Company | 🔵 Month 6-12 |
Buzz was built from scratch by Ogie — a 20+ year Executive Chef who transitioned to crypto BD. No CS degree. No engineering team. Just conversational collaboration with Claude (Opus 4.6), building one feature at a time from a phone in Indonesia during Ramadan.
From first deploy (Feb 1, 2026) to Alpha Buzz (v7.3.1a) in 39 days. Three-layer architecture: Foundation → Strategic Orchestrator → Alpha Intelligence. Self-improving agent that learns from every scan, every score, every outreach.
The formula: Hermes Brain + Buzz Body = Alpha Model.
Total infrastructure cost: ~$12.72/month. Revenue target: $5,000/listing.
| Platform | URL |
|---|---|
| SolCex Exchange | solcex.io |
| Docs | docs.solcex.cc |
| @BuzzBySolCex | |
| SolCex Twitter | @SolCex_Exchange |
| Telegram Bot | @BuzzBySolCex_bot |
| Builder | @hidayahanka1 |
| linkedin.com/in/howtobecomeachef |
Proprietary — SolCex Exchange. All rights reserved.
🐝 "Identity first. Intelligence deep. Commerce autonomous. Cost disciplined. Strategy embedded. Memory persistent. Learning continuous."
Buzz BD Agent v7.3.1a — Alpha Buzz (Phase 0 Complete) | Sprint Day 21 | Mar 11, 2026 | Jakarta, Indonesia