lossless-claw

XmetaV -- OpenClaw Command Center

> Your central hub for managing OpenClaw agents, gateways, and infrastructure on Mac Studio (M3 Ultra)

Last updated: 2026-03-08 | OpenClaw 2026.3.2 | XmetaV Command Center v28.2

 ___   ___                    __           ___   ___
 \  \ /  / _ __ ___    ___  _/  |_  __ _   \  \ /  /
  \  V  / | '_ ` _ \  / _ \ \   __\/ _` |   \  V  /
  /  X  \ | | | | | ||  __/  |  | | (_| |   /     \
 /__/ \__\|_| |_| |_| \___|  |__|  \__,_|  /__/ \__\

      [ COMMAND CENTER : AGENT ORCHESTRATION ]
  _______________________________________________
 |                                               |
 |   agents:  14 (+ dynamic)                     |
 |   skills:  12 ethskills (wallets/tools/l2s..) |
 |   swarm:   parallel | pipeline | collab       |
 |   payments: x402 USDC micro-payments (Base)   |
 |   cross:    Base↔Solana bridge+Jupiter+Kamino |
 |   identity: ERC-8004 NFT #16905 (Base)        |
 |   token:    $XMETAV ERC-20 (Base)             |
 |   dashboard: Next.js + Supabase (cyberpunk)   |
 |   models:  kimi-k2.5:cloud (256k, all agents) |
 |   gateway: ws://127.0.0.1:18789              |
 |   engine:  Ollama  |  M3 Ultra (60 GPU cores) |
 |_______________________________________________|

---

Features

• Control Plane Dashboard — Cyberpunk-themed Next.js web UI for agent chat, fleet management, swarm orchestration, and bridge control (Vercel-deployable)
• XMETAV HQ (Arena) — Isometric office visualization with PixiJS: boss office, meeting table with seat-based agent meetings, workstations, glowing orb avatars, real-time command pulses, dispatch beams, and reactive holo screens -- all driven by live Supabase events with periodic sync
• Cyberpunk Arena Evolution — Neon aura rings with chromatic aberration on avatars, hex rain data torrents for streaming, RGB-split failure glitches, holographic meeting ring with scanlines, distant city skyline parallax, and perspective neon grid floor
• Neural Swarm Links — Real-time neural data links between agents during swarm runs: pipeline chains, hub-and-spoke topology, per-agent status rings, flowing data particles, chromatic ghost lines — all driven by Supabase Realtime on swarm_runs/swarm_tasks
• Live Log Streaming — Real-time log viewer with severity filters, agent filters, search, and auto-scroll
• Swarm Dashboard — Create, monitor, and review multi-agent swarm runs from the browser with live streaming output
• Agent Factory — main agent can create new agents, scaffold apps, create GitHub repos, and manage the fleet
• Swarm Orchestration — parallel, pipeline, and collaborative multi-agent task execution (CLI + dashboard)
• Fleet Controls — Enable/disable agents from the dashboard with bridge-side enforcement
• x402 Payments — Autonomous USDC micro-payments on Base via @x402/express + CDP facilitator with JWT auth (pay-per-use API gating: $0.01-$3.00 per endpoint, includes trade execution with %-of-capital pricing)
• ERC-8004 Identity — On-chain agent identity (NFT #16905) with x402Support metadata, identity resolution middleware, and /agent/:id/payment-info discovery endpoint
• Voice Commands — Speak to agents via Whisper STT + TTS with x402 payment gating ($0.05-$0.08 per request)
• Persistent Agent Memory Bus — Supabase-backed memory entries (_shared + per-agent) injected by the bridge at dispatch time, with outcome capture after completion (complements OpenClaw session history)
• Soul Agent (Memory Orchestrator) — Dedicated memory orchestration agent with dream consolidation, association building, context packet curation, fleet-wide memory retrieval learning, and surveillance desk with mini fleet monitors in the Arena
• Consciousness Tab — Dual-aspect awareness dashboard: unified Main↔Soul split view, force-directed memory graph, on-chain anchor timeline, context metrics, dream mode status, and mini arena — all with 30s auto-refresh from Supabase
• Dreamscape Visualization — Canvas-based dream session renderer: hexagonal memory shards with glow, diamond insight crystals by category, curved synaptic bridges with flowing data particles, mouse-interactive repulsion, ambient data motes, and insight ticker strip
• Memory Crystal System (Materia) — Final-Fantasy-inspired memory gamification: living crystals with XP/levels/star ratings (1-6★), class evolution (anchor→godhand), fusion recipes (FF7-style), memory summons with animated rituals, limit breaks spawning legendary 6★ crystals, explorable Memory Cosmos world map, achievements, and daily quests
• Memory Cosmos World — Pannable/zoomable explorable memory landscape at /memory-cosmos: golden-spiral island layout with terrain types (city/wasteland/forest), neon highway bridges with data particles, crystal-topped islands, star parallax background, hover info panels
• Swarm Network View — Canvas-based neural network topology on the Consciousness page: circular node layout with agent colors, curved links with chromatic aberration, pipeline arrows, data flow particles, run status ticker, and empty-state handling
• Cross-Chain Swaps (Base↔Solana) — Full multi-chain pipeline: USDC bridge (Base→Solana via CCTP), Jupiter Ultra swaps (SOL/BONK/JUP), Kamino vault deposits for yield, return bridge (Solana→Base). Batch queue for sub-threshold amounts, job lifecycle tracking in Supabase, x402 payment gating ($0.65/swap)
• Jupiter Ultra Integration — RPC-less token swaps on Solana via Jupiter Ultra API with multi-route aggregation (PancakeSwap, Whirlpool, Meteora DLMM, TesseraV), automatic slippage protection, price impact safety checks
• Kamino Yield Vaults — SDK-first via @kamino-finance/klend-sdk: vault data (APY, holdings, exchange rate), user positions, deposit/withdraw with @solana/kit compat. REST API fallback
• Kamino Borrow/Lend — Full lending market integration via klend-sdk: KaminoMarket overview, user obligations with LTV tracking, deposit-collateral/borrow/repay/withdraw-collateral via KaminoAction. 5 gated + 3 free endpoints
• Standalone Kamino Endpoints — Direct POST /kamino/deposit ($0.15) and POST /kamino/withdraw ($0.15) for vault operations without a full cross-chain swap
• Trading/DeFi Dashboard — Dedicated /trading page with 3-column layout: CrossChainPanel (queue stats, swap quotes), KaminoPanel (vault overview, deposit/withdraw), KaminoBorrowPanel (market TVL, reserves, obligations, 4-action form)
• Swap Execution — Agent-initiated token swaps with gas/balance pre-checks, voice normalization (spoken aliases → canonical symbols), and swap history tracking via agent_swaps table
• Streaming Pipeline v2 — 2.5× faster response rendering: chunk size 160, flush 80ms, token batching (6/15ms), RAF-aligned 50ms throttle, React.memo StreamingBubble
• $XMETAV Token — ERC-20 on Base Mainnet (0x5b56CD209e3F41D0eCBf69cD4AbDE03fC7c25b54) with tiered discounts (10-50% off) on x402 endpoints
• EthSkills Integration — 12 blockchain/Ethereum skills from ethskills.com installed across fleet agents: wallets (main), tools/l2s/orchestration/addresses/concepts/security/standards/frontend-ux/frontend-playbook/building-blocks (web3dev), gas/l2s (oracle), gas/standards/addresses/concepts (midas). Skills displayed as badges in Fleet table and Identity page
• Multi-agent management (14 agents + dynamic): main, sentinel, soul, briefing, oracle, alchemist, midas, web3dev, akua, akua_web, basedintern, basedintern_web, vox, scholar
• Multi-model support (local qwen2.5 + cloud kimi-k2.5:cloud with 256k context)
• App scaffolding (Node.js, Python, Next.js, Hardhat, bots, FastAPI)
• GitHub integration for automated repo creation and pushing
• Full tool calling (exec, read, write, process, browser, web, cron, gateway, sessions)
• OpenClaw-managed browser automation
• Ollama integration with Apple Silicon GPU acceleration (M3 Ultra, 60 GPU cores)

---

What is XmetaV?

XmetaV is your operational command center for managing OpenClaw -- an AI agent automation platform. This repository contains:

• Control Plane Dashboard -- Cyberpunk web UI (Next.js + Supabase) for agent chat, fleet management, swarm orchestration, and real-time monitoring
• Bridge Daemon -- Local Node.js service that bridges the dashboard to OpenClaw CLI via Supabase Realtime
• Agent Factory -- Create agents on the fly, scaffold apps, create GitHub repos, manage the fleet
• Swarm Engine -- Orchestrate multi-agent tasks (parallel, pipeline, collaborative) via CLI or dashboard
• Setup & Fix Scripts -- Automated solutions for common issues
• Configuration Templates -- Battle-tested configs for Ollama + Kimi K2.5
• Documentation -- Runbooks, checklists, and troubleshooting guides
• x402 Payment Service -- Express server gating XmetaV endpoints with USDC micro-payments on Base + cross-chain swap engine (Base↔Solana)
• ERC-8004 Identity -- On-chain agent NFT (identity + reputation) on Base mainnet
• Agent Definitions -- Multi-agent profiles and workspaces
• Infrastructure as Code -- Reproducible OpenClaw deployments

---

Repository Structure

XmetaV/
|-- README.md                 # You are here
|-- LICENSE                   # MIT License
|
|-- dashboard/                # Control Plane Dashboard (Next.js + Supabase)
|   |-- src/
|   |   |-- app/
|   |   |   |-- (dashboard)/  # Protected routes (Command Center, Agent Chat, Swarms, Fleet)
|   |   |   |-- arena/        # XMETAV HQ isometric office visualization (standalone)
|   |   |   |-- auth/         # Login page
|   |   |   +-- api/          # API routes (commands, swarms, agents, bridge)
|   |   |-- components/       # UI: Sidebar, AgentChat, FleetTable, SwarmCreate, Consciousness, Crystals, etc.
|   |   |   +-- arena/        # PixiJS renderers (iso, background, office, avatars, effects)
|   |   |-- hooks/            # Realtime hooks (messages, bridge, sessions, swarms, consciousness, crystals)
|   |   +-- lib/              # Supabase clients, types
|   |-- bridge/               # Bridge Daemon (Node.js)
|   |   |-- src/              # executor, swarm-executor, streamer, heartbeat
|   |   +-- lib/              # openclaw CLI wrapper, Supabase client, x402 client
|   |-- x402-server/          # x402 payment-gated Express service + cross-chain swap engine
|   |-- erc8004/              # ERC-8004 agent identity (registration, ABIs, client)
|   |-- scripts/              # DB migrations (setup-db*.sql)
|   +-- README.md             # Dashboard documentation
|
|-- scripts/                  # Executable automation scripts
|   |-- openclaw-fix.sh       # Main fix script (gateway + ollama + locks)
|   |-- start-gateway.sh      # Start gateway in background
|   |-- stop-all.sh           # Stop processes + clear stale locks
|   |-- health-check.sh       # Quick system health verification
|   |-- agent-task.sh         # Single atomic task wrapper
|   |-- agent-pipeline.sh     # Multi-step pipeline workflows
|   |-- create-agent.sh       # Agent Factory — create agents
|   |-- build-app.sh          # Agent Factory — scaffold apps
|   |-- manage-agents.sh      # Agent Factory — manage fleet
|   +-- swarm.sh              # Swarm — multi-agent orchestration (CLI)
|
|-- configs/                  # Configuration files & templates
|   +-- openclaw.json.fixed   # Known-good config for macOS + Ollama
|
|-- templates/                # Agent identity & swarm templates
|   |-- agents/               # Per-template identity files
|   |   |-- general.md        # Generic agent template
|   |   |-- coding.md         # Software development agent
|   |   |-- bot.md            # Discord/Telegram bot agent
|   |   |-- research.md       # Web research agent
|   |   +-- devops.md         # Infrastructure/ops agent
|   +-- swarms/               # Pre-built swarm manifests
|       |-- health-all.json   # Parallel health check
|       |-- ship-all.json     # Parallel build+test
|       |-- research-implement.json  # Pipeline: research -> implement
|       +-- code-review.json  # Collaborative code review
|
|-- capabilities/             # Quick-reference command guides
|   |-- README.md             # Capabilities overview
|   |-- quick-commands.md     # Essential daily-use commands
|   |-- agent-tasks.md        # AI agent usage examples
|   |-- cheatsheet.md         # One-page reference card
|   |-- management.md         # System administration commands
|   |-- expand.md             # How to add models, skills, channels, agents
|   |-- x402-payments.md      # x402 autonomous payment protocol reference
|   +-- erc8004-identity.md   # ERC-8004 on-chain agent identity reference
|
    +-- docs/                     # Documentation & runbooks
    |-- ARCHITECTURE.md       # System architecture overview
    |-- AGENTS.md             # Agent configuration guide
    |-- STATUS.md             # Current known-good settings + checks
    |-- TROUBLESHOOTING.md    # Common issues & solutions
    |-- OLLAMA-SETUP.md       # Ollama integration guide
    |-- OPENCLAW-FIX-CHECKLIST.md  # Verification checklist
    |-- GITHUB-SKILL-STATUS.md     # GitHub skill status
    |-- SWARM.md              # Multi-agent swarm reference
    +-- agents/               # Per-agent runbooks
        |-- README.md
        |-- main.md           # main agent runbook
        |-- basedintern.md    # basedintern agent runbook
        |-- akua.md           # akua agent runbook
        |-- dynamic.md        # Dynamic agent runbook
        |-- briefing.md       # briefing (context curator) runbook
        |-- oracle.md         # oracle (on-chain intel) runbook
        |-- alchemist.md      # alchemist (tokenomics) runbook
        |-- midas.md          # midas (revenue & growth) runbook
        |-- web3dev.md        # web3dev (blockchain dev) runbook
        |-- sentinel.md      # sentinel (fleet ops) runbook

---

Quick Start

Prerequisites

| Requirement | Version | Check Command | |-------------|---------|---------------| | OpenClaw CLI | 2026.2.17+ | openclaw --version | | Node.js | 25.x | node --version | | Ollama | 0.17.4+ (macOS app) | ollama --version | | just | 1.46+ | just --version | | pnpm | 10.x | pnpm --version | | macOS | 26.x (Apple Silicon) | sw_vers |

> Platform: Mac Studio M3 Ultra (96GB, 28 CPU cores, 60 GPU cores). Ollama runs as a macOS app with Apple Silicon GPU acceleration — no CUDA needed.

1. Clone & Setup

git clone https://github.com/youruser/XmetaV.git
cd XmetaV
chmod +x scripts/*.sh

2. Fix Common Issues (First Run)

./scripts/openclaw-fix.sh

This script will:

• Kill stale processes
• Remove lock files
• Patch configuration for local Ollama
• Start the gateway on port 18789
• Verify everything works

3. Verify Installation

openclaw health

# Use --local flag for reliable agent calls (bypasses gateway websocket)
openclaw agent --agent main --local --thinking off \
  --session-id test_$(date +%s) --message "What is 2+2? Reply with just 4."

> Note: The --local flag runs the agent embedded (bypasses gateway websocket). This is the recommended mode for local Ollama usage.

---

Available Scripts

| Script | Purpose | Usage | |--------|---------|-------| | openclaw-fix.sh | Complete fix -- kills zombies, clears locks, patches config, starts gateway | ./scripts/openclaw-fix.sh | | start-gateway.sh | Start gateway in background on port 18789 | ./scripts/start-gateway.sh | | stop-all.sh | Stop all OpenClaw processes | ./scripts/stop-all.sh | | health-check.sh | Quick health verification | ./scripts/health-check.sh | | agent-task.sh | Single atomic task (fresh session, anti-stall) | ./scripts/agent-task.sh basedintern "task" | | agent-pipeline.sh | Multi-step workflows (health, ship, fix, evolve) | ./scripts/agent-pipeline.sh health | | create-agent.sh | Agent Factory -- create new agents | ./scripts/create-agent.sh --id myagent | | build-app.sh | Agent Factory -- scaffold apps | ./scripts/build-app.sh --type node --workspace /path | | manage-agents.sh | Agent Factory -- manage agent fleet | ./scripts/manage-agents.sh list | | swarm.sh | Swarm -- multi-agent orchestration (parallel, pipeline, collab) | ./scripts/swarm.sh --parallel ... | | briefing-agent.sh | Briefing Agent -- health sentinel + auto-fix + distill + sitrep | ./scripts/briefing-agent.sh | | distill.sh | Memory Distill -- consolidate activity into MEMORY.md + refresh SITREP | ./scripts/distill.sh | | oracle-agent.sh | Oracle Agent -- on-chain intel, gas, prices, sentiment, alerts | ./scripts/oracle-agent.sh | | alchemist-agent.sh | Alchemist Agent -- $XMETAV tokenomics, emissions, staking curves | ./scripts/alchemist-agent.sh | | midas-agent.sh | Midas Agent -- x402 revenue analytics, pricing, growth pipeline | ./scripts/midas-agent.sh | | web3dev-agent.sh | Web3Dev Agent -- compile, test, audit, deploy smart contracts | ./scripts/web3dev-agent.sh |

---

Configuration

State Directory

This repo uses the default OpenClaw config (no profile flag needed):

| Setting | Value | |---------|-------| | State Directory | ~/.openclaw/ | | Config File | ~/.openclaw/openclaw.json | | Gateway Port | 18789 | | Gateway Mode | local |

Model Provider: Ollama

| Setting | Value | |---------|-------| | Base URL | http://127.0.0.1:11434/v1 | | API Mode | openai-responses | | API Key | "local" (required placeholder) |

Available models:

| Model | Type | Context Window | |-------|------|----------------| | qwen2.5:7b-instruct | Local | 32,768 | | kimi-k2.5:cloud | Cloud (Ollama) | 262,144 (256k) |

> Why openai-responses? It's required for tool calling (exec/read/write/process). If you only want chat (no tools), openai-completions can work but won't inject tool schemas.

Ollama Cloud limits (Kimi)

Cloud models (like kimi-k2.5:cloud) are subject to plan/session usage limits. If you hit the quota you'll see HTTP 429. Fix: wait for reset or upgrade the plan. Cloud auth uses ollama signin.

Key Config Values

{
  "gateway": {
    "port": 18789,
    "mode": "local",
    "bind": "loopback"
  },
  "models": {
    "providers": {
      "ollama": {
        "baseUrl": "http://127.0.0.1:11434/v1",
        "api": "openai-responses",
        "apiKey": "local"
      }
    }
  },
  "agents": {
    "defaults": {
      "model": {
        "primary": "ollama/qwen2.5:7b-instruct"
      }
    }
  }
}

---

Agents

Agent: main (default)

| Property | Value | |----------|-------| | ID | main | | Model | ollama/kimi-k2.5:cloud (256k context) | | Workspace | ~/.openclaw/workspace | | Tools | full (fs, runtime, web, browser, sessions, automation) | | Skills | wallets (EOAs, Safe multisig, EIP-7702, ERC-4337) | | Role | Orchestrator — agent factory + swarm + command center |

openclaw agent --agent main --local --thinking off --message "Hello!"

Agent: briefing (context curator)

| Property | Value | |----------|-------| | ID | briefing | | Model | kimi-k2.5:cloud (256k context) | | Workspace | /home/manifest/briefing | | Tools | coding (exec, read, write) | | Role | Context Curator -- continuity, health sentinel, memory distillation |

The morning person who has the coffee ready. Maintains SITREP.md (live situation report) and MEMORY.md (long-term memory) so main doesn't waste sessions re-discovering context. Runs on a cron schedule with auto-fix for common health issues.

# Run manually
./scripts/briefing-agent.sh

# Or via cron (every hour):
# 0 * * * * /home/manifest/XmetaV/scripts/briefing-agent.sh >> /tmp/briefing-agent.log 2>&1

See docs/agents/briefing.md for full documentation.

Agent: oracle (on-chain intelligence)

| Property | Value | |----------|-------| | Purpose | On-chain intelligence and market sentinel | | Workspace | /home/manifest/oracle | | Tools | coding (exec, read, write) | | Model | ollama/kimi-k2.5:cloud | | Skills | gas (tx costs, L1 vs L2), l2s (Arbitrum, Optimism, Base, zkSync) | | Role | Monitor gas, prices, chain activity, protocol intel, crypto sentiment |

Provides real-time market intelligence via public APIs (CoinGecko, Etherscan, DeFiLlama, CryptoCompare). Writes ORACLE.md reports and surfaces alerts before anyone has to ask.

# Run manually
./scripts/oracle-agent.sh

# Quick alerts every 15 min, full report every hour:
# */15 * * * * /home/manifest/XmetaV/scripts/oracle-agent.sh --alerts >> /tmp/oracle-alerts.log 2>&1
# 0 * * * *   /home/manifest/XmetaV/scripts/oracle-agent.sh --report >> /tmp/oracle-report.log 2>&1

See docs/agents/oracle.md for full documentation.

Agent: alchemist (tokenomics)

| Property | Value | |----------|-------| | Purpose | $XMETAV tokenomics specialist | | Workspace | /home/manifest/alchemist | | Tools | coding (exec, read, write) | | Model | ollama/kimi-k2.5:cloud | | Role | Supply tracking, emission modeling, holder analysis, staking curves, liquidity intel |

Reads directly from the $XMETAV contract on Base Mainnet. Models emission schedules, identifies sell pressure windows, and recommends staking parameters. Writes TOKENOMICS.md reports.

# Run manually
./scripts/alchemist-agent.sh

# Health check every 6h, full report daily at 8 AM:
# 0 */6 * * * /home/manifest/XmetaV/scripts/alchemist-agent.sh --health >> /tmp/alchemist.log 2>&1
# 0 8 * * *   /home/manifest/XmetaV/scripts/alchemist-agent.sh --report >> /tmp/alchemist.log 2>&1

See docs/agents/alchemist.md for full documentation.

Agent: midas (revenue & growth)

| Property | Value | |----------|-------| | Purpose | Revenue intelligence & growth strategy | | Workspace | /home/manifest/midas | | Tools | coding (exec, read, write) | | Model | ollama/kimi-k2.5:cloud | | Skills | revenue, pricing, growth, forecast | | Role | x402 analytics, endpoint pricing, revenue forecasts, growth pipeline, R&D prioritisation |

Sits between Oracle and Alchemist (arena seat 165°). Synthesizes market intel + tokenomics into revenue strategy. Tracks all x402 payment flows, analyzes endpoint performance, recommends pricing adjustments, and maintains a ROI-ranked growth opportunity pipeline.

# Run manually
./scripts/midas-agent.sh

# Revenue snapshot daily at 6 AM, endpoints every 12h, growth scan Monday 8 AM:
# 0 6 * * *    /home/manifest/XmetaV/scripts/midas-agent.sh --report    >> /tmp/midas.log 2>&1
# 0 */12 * * * /home/manifest/XmetaV/scripts/midas-agent.sh --endpoints >> /tmp/midas.log 2>&1
# 0 8 * * 1   /home/manifest/XmetaV/scripts/midas-agent.sh --growth    >> /tmp/midas.log 2>&1

See docs/agents/midas.md for full documentation.

Agent: web3dev (blockchain developer)

| Property | Value | |----------|-------| | Purpose | Dedicated blockchain developer — compile, test, audit, deploy | | Workspace | /home/manifest/web3dev | | Tools | coding (exec, process, read, write) | | Model | ollama/kimi-k2.5:cloud | | Skills | tools, l2s, orchestration, addresses, concepts, security, standards, frontend-ux, frontend-playbook, building-blocks (10 ethskills) | | Role | Smart contract dev, security auditor, deployment engineer, x402 maintainer |

Owns all Hardhat projects (Akua 18 contracts, $XMETAV token, BasedIntern). Static security audits, gas/size analysis, and production-ready contract scaffolding (ERC-20, staking, vesting, escrow).

# Run manually
./scripts/web3dev-agent.sh

# Status every 12h, weekly audit:
# 0 */12 * * * /home/manifest/XmetaV/scripts/web3dev-agent.sh --status >> /tmp/web3dev.log 2>&1
# 0 6 * * 1    /home/manifest/XmetaV/scripts/web3dev-agent.sh --report >> /tmp/web3dev.log 2>&1

See docs/agents/web3dev.md for full documentation.

Agent: sentinel (fleet lifecycle)

| Property | Value | |----------|-------| | Purpose | Agent lifecycle manager — spawn coordination, fleet health | | Workspace | /home/manifest/sentinel | | Tools | coding (exec, read, write) | | Model | ollama/kimi-k2.5:cloud | | Role | Spawn coordinators, resource managers, inter-agent communication, fleet health monitoring |

Manages agent lifecycles and fleet health. Coordinates agent spawning, monitors resource usage, handles inter-agent communication routing, and maintains fleet health dashboards. Commands: status, health, spawn, queue, errors.

See docs/agents/sentinel.md for full documentation.

Agent: vox (brand & campaigns)

| Property | Value | |----------|-------| | Purpose | Brand voice & X campaign specialist — content strategy, voice calibration, competitor analysis | | Workspace | ~/.openclaw/agents/vox | | Tools | coding (exec, read, write) | | Model | ollama/kimi-k2.5:cloud | | Skills | content-strategy, voice-calibration, competitor-analysis | | Room | OPS Outreach | | Color | Cyan (#06b6d4) |

Generates tweet threads, audits brand voice consistency, monitors competitor messaging, and creates content calendars. Integrates with midas (revenue milestones), oracle (market events), briefing (SITREP recaps), and soul (engagement memory). Commands: campaign, voice --audit, calendar, competitor, crisis, report, health.

# Generate a campaign thread
./scripts/agent-task.sh vox "Create a thread about our memory anchoring system"

# Voice consistency audit
./scripts/vox-agent.sh --voice

# Weekly content calendar
./scripts/vox-agent.sh --calendar

See docs/agents/vox.md for full documentation.

Agent: basedintern (coding) + basedintern_web (full)

| Property | basedintern | basedintern_web | |----------|---------------|-------------------| | Model | kimi-k2.5:cloud (256k) | kimi-k2.5:cloud (256k) | | Workspace | /home/manifest/basedintern | /home/manifest/basedintern | | Tools | coding (exec, read, write, process) | full (all tools + browser + web) | | Use for | 90% of work (code, tests, commits) | Only browser/web automation |

Why two agents? The coding profile advertises ~4 tools (small schema). The full profile advertises 20+ tools. Fewer tools = fewer tokens per Kimi call = faster + less 429s.

# Default (lean, fast)
openclaw agent --agent basedintern --local --thinking off \
  --session-id bi_$(date +%s) --message "Run npm test."

# Full tools (only when needed)
openclaw agent --agent basedintern_web --local --thinking off \
  --session-id biweb_$(date +%s) --message "Use web_fetch to check a URL."

Agent: akua (coding) + akua_web (full)

| Property | akua | akua_web | |----------|--------|------------| | Model | kimi-k2.5:cloud (256k) | kimi-k2.5:cloud (256k) | | Workspace | /home/manifest/akua | /home/manifest/akua | | Tools | coding (exec, read, write, process) | full (all tools + browser + web) | | Use for | 90% of work (contracts, tests, commits) | Only browser/web automation | | Repo | Metavibez4L/akua | Same |

# Default (lean, fast)
openclaw agent --agent akua --local --thinking off \
  --session-id akua_$(date +%s) --message "Run /repo-ops compile."

# Full tools (only when needed)
openclaw agent --agent akua_web --local --thinking off \
  --session-id akuaweb_$(date +%s) --message "Use web_fetch to check a URL."

Creating New Agents (Agent Factory)

The main agent can create agents autonomously, or you can use the scripts directly:

# Create a new agent with workspace, identity files, and config entry
./scripts/create-agent.sh --id researcher \
  --template research \
  --description "Web research and data gathering" \
  --web  # also create researcher_web companion

# Create agent + auto-create a GitHub repo and push
./scripts/create-agent.sh --id researcher \
  --template research --web \
  --github --private  # creates Metavibez4L/researcher on GitHub

# Scaffold a project in the agent's workspace
./scripts/build-app.sh --type node --workspace /home/manifest/researcher

# Scaffold + push to GitHub
./scripts/build-app.sh --type node --workspace /home/manifest/researcher --github

# Check the fleet
./scripts/manage-agents.sh list

# Run it
./scripts/agent-task.sh researcher "What can you do?"

Or ask the main agent to do it:

openclaw agent --agent main --local \
  --message "Create a Discord bot agent called social-bot and scaffold a bot project for it"

Templates: coding, bot, research, devops, general

App types: node, python, nextjs, hardhat, bot, fastapi, script

See docs/agents/dynamic.md for full documentation.

Control Plane Dashboard

A cyberpunk-themed web dashboard for managing the entire XmetaV ecosystem from a browser.

# Run locally
cd dashboard && npm install && npm run dev
# Open http://localhost:3000

# Start the bridge daemon (connects dashboard to OpenClaw CLI)
cd dashboard/bridge && npm install && npm start

Dashboard Pages:

| Page | Description | |------|-------------| | / | Command Center -- bridge health, fleet summary, command history, quick command | | /agent | Agent Chat -- full-screen streaming chat with agent selector | | /swarms | Swarms -- create, monitor, and review multi-agent swarm runs | | /fleet | Fleet -- agent status table with enable/disable toggles and task dispatch | | /payments | Payments -- x402 wallet status, spend tracking, payment history, gated endpoints | | /identity | Identity -- ERC-8004 on-chain agent identity, reputation, and NFT details | | /token | $XMETAV -- ERC-20 token balance, tier status, discount table, holder benefits | | /consciousness | Consciousness -- Dual-aspect awareness: memory graph, anchor timeline, context metrics, dream mode, mini arena | | /memory-cosmos | Memory Cosmos -- Crystal materia inventory, fusion chamber, summon overlay, limit breaks, explorable world, achievements, quests | | /arena | XMETAV HQ -- Isometric office visualization with live agent activity (PixiJS) | | /logs | Live Logs -- Real-time log streaming with severity/agent filters and search | | /trading | Trading / DeFi -- Cross-chain queue stats, swap quote tool, Kamino vault deposit/withdraw |

Key Features:

• Swarm Dashboard -- Create swarms from templates or custom builder, "Let Main Agent Decide" button, live progress bars, per-task streaming output, run history with filters
• Agent Controls -- Enable/disable agents from the Fleet page; disabled agents have commands blocked by the bridge
• Bridge Controls -- Start/stop the local bridge daemon from the dashboard
• Real-time -- All data updates live via Supabase Realtime (no polling)
• Cyberpunk UI -- Neon blue/dark hacker aesthetic with glitch effects, scanlines, and animated elements
• Memory Crystal System -- FF-style crystal materia with XP, fusion, summons, limit breaks, and an explorable Memory Cosmos world

Architecture: Browser (Vercel) <-> Supabase (command bus) <-> Bridge Daemon (WSL) <-> OpenClaw CLI <-> Agents

See dashboard/README.md for full documentation.

Swarm Orchestration

Dispatch tasks across multiple agents with three modes (CLI or dashboard):

# Parallel: run independent tasks simultaneously
./scripts/swarm.sh --parallel \
  basedintern "Run npm test" \
  akua "Run /repo-ops compile"

# Pipeline: chain agents, output flows forward
./scripts/swarm.sh --pipeline \
  main "Research error handling best practices" \
  basedintern "Apply the findings to the codebase"

# Collaborative: multiple perspectives on the same problem
./scripts/swarm.sh --collab "Review security of our auth flow" \
  basedintern akua

# Pre-built templates
./scripts/swarm.sh templates/swarms/health-all.json

# Check results
./scripts/swarm.sh --status
./scripts/swarm.sh --results <run-id>

Or let the main agent orchestrate swarms via its Swarm skill:

openclaw agent --agent main --local \
  --message "Run a parallel health check across all repo agents"

See docs/SWARM.md for full documentation.

---

System Architecture

flowchart TB
    subgraph XMETAV["XmetaV (This Repo)"]
        direction LR
        SCRIPTS["Scripts"] ~~~ CONFIGS["Configs"] ~~~ DOCS["Docs"] ~~~ TEMPLATES["Templates"]
    end

    subgraph RUNTIME["OpenClaw Runtime"]
        GW["Gateway — ws://127.0.0.1:18789"]
        subgraph GWSVC[" "]
            direction LR
            AR["Agent\nRuntime"]
            SM["Session\nManager"]
            CR["Channel\nRouter"]
            SE["Skill\nExecutor"]
        end

        subgraph ORCH["main agent (ORCHESTRATOR)"]
            direction TB
            subgraph FAC["Agent Factory"]
                direction LR
                CA["create-agent.sh"]
                BA2["build-app.sh"]
                MG["manage-agents.sh"]
            end
            subgraph SWE["Swarm Engine — swarm.sh"]
                direction LR
                S_P["Parallel\n⫘ A B C"]
                S_PI["Pipeline\nA → B → C"]
                S_C["Collaborative\nA + B → synth"]
            end
        end

        subgraph FLEET["Agent Fleet (10+)"]
            direction LR
            F_MAIN["main\n(orch.)"]
            F_INTEL["sentinel\nbriefing\noracle\nalchemist"]
            F_DEV["web3dev\nakua (+web)\nbasedintern (+web)"]
            F_DYN["dynamic\nagents"]
        end
    end

    subgraph PROV["Model Providers"]
        direction LR
        OLL["Ollama (Local)\nhttp://127.0.0.1:11434\nqwen2.5 · kimi-k2.5:cloud"]
        CLD["Cloud Providers\nAnthropic · OpenAI"]
    end

    subgraph EXT["External"]
        GH["GitHub\nMetavibez4L"]
    end

    subgraph BASE["Base Mainnet"]
        direction LR
        X402P["x402 USDC\nPayments"]
        ERC["ERC-8004\nIdentity #16905"]
    end

    XMETAV ==> GW
    GW --> GWSVC --> ORCH
    ORCH --> FLEET
    FAC -.->|--github| GH
    FLEET ==> PROV
    FLEET -.->|pay / identify| BASE

    style XMETAV fill:#1a1a2e,stroke:#e94560,color:#fff
    style RUNTIME fill:#16213e,stroke:#e94560,color:#fff
    style ORCH fill:#0f3460,stroke:#e94560,color:#fff
    style FAC fill:#1a1a4e,stroke:#16c79a,color:#fff
    style SWE fill:#1a1a4e,stroke:#f7b731,color:#fff
    style FLEET fill:#1a1a3e,stroke:#a29bfe,color:#fff
    style PROV fill:#222,stroke:#888,color:#fff
    style EXT fill:#161b22,stroke:#58a6ff,color:#fff
    style GH fill:#161b22,stroke:#58a6ff,color:#fff
    style BASE fill:#0052ff,stroke:#fff,color:#fff

---

Common Commands

Gateway Management

# Start gateway (background)
./scripts/start-gateway.sh

# Check gateway status
openclaw health

# View gateway logs
tail -f ~/.openclaw/gateway.log

Browser Automation (OpenClaw-managed)

openclaw browser start
openclaw browser open https://example.com
openclaw browser snapshot
openclaw browser click e123

Agent Operations

# Simple message (default agent)
openclaw agent --message "What is 2+2?"

# With specific agent and session
openclaw agent --agent basedintern \
  --session-id my-session \
  --message "Run npm test"

# List sessions
openclaw sessions list

Configuration

# View current config
openclaw config get

# Set a value
openclaw config set gateway.mode local

# View specific key
openclaw config get models.providers.ollama.api

---

Troubleshooting

| Issue | Solution | |-------|----------| | Gateway closed (1006) | Run ./scripts/openclaw-fix.sh -- gateway not running or wrong port | | Waiting for agent reply forever | Use --local --thinking off, clear stale locks, ensure api=openai-responses | | Session locked | find ~/.openclaw -name "*.lock" -delete | | Connection refused to Ollama | ollama serve or check systemd status | | Port 18789 already in use | fuser -k 18789/tcp then restart gateway | | No API key found for provider ollama | openclaw config set models.providers.ollama.apiKey "local" | | Browser start fails | Install browser deps (see docs/STATUS.md) |

See docs/TROUBLESHOOTING.md for detailed solutions.

---

Documentation Index

| Document | Description | |----------|-------------| | dashboard/README.md | Control Plane Dashboard — setup, architecture, pages, bridge | | ARCHITECTURE.md | System architecture deep-dive (includes dashboard) | | AGENTS.md | Agent configuration, tool profiles, and customization | | agents/ | Per-agent runbooks (main, briefing, oracle, alchemist, web3dev, basedintern, akua, dynamic) | | STATUS.md | Current known-good settings + verification commands | | SWARM.md | Multi-agent swarm orchestration reference | | TROUBLESHOOTING.md | Common issues & solutions (includes dashboard, x402, voice) | | OLLAMA-SETUP.md | Ollama integration guide | | OPENCLAW-FIX-CHECKLIST.md | Verification checklist | | GITHUB-SKILL-STATUS.md | GitHub skill status | | capabilities/x402-payments.md | x402 autonomous payment protocol reference | | capabilities/erc8004-identity.md | ERC-8004 on-chain agent identity reference | | capabilities/voice-commands.md | Voice command & response system reference | | docs/memory/ | Memory system architecture — anchoring, Soul orchestration | | capabilities/quick-commands.md | Essential daily-use commands |

---

GitHub Skill Integration

The GitHub skill is installed, authenticated, and working with OpenClaw agents.

• To use: /github help, /github status, /github issue list, etc.
• Requires: GitHub CLI (gh) installed and authenticated (gh auth login).
• See docs/GITHUB-SKILL-STATUS.md for status.

---

Development

Adding New Scripts

1. Create script in scripts/ 2. Make executable: chmod +x scripts/your-script.sh 3. Add documentation in script header 4. Update this README

Configuration Changes

1. Test with openclaw config set ... 2. Export working config to configs/openclaw.json.fixed 3. Document changes in docs/

---

On-Chain Contracts (Base Mainnet)

All contracts are deployed on Base Mainnet (chain ID 8453, eip155:8453).

| Contract | Address | Description | |----------|---------|-------------| | $XMETAV Token | 0x5b56CD209e3F41D0eCBf69cD4AbDE03fC7c25b54 | ERC-20 token (1B fixed supply) — tiered discounts on x402 | | ERC-8004 Identity | 0x8004A169FB4a3325136EB29fA0ceB6D2e539a432 | IdentityRegistry — Agent NFT #16905 | | ERC-8004 Reputation | 0x8004b1041543F0eB1f3459E8a2FC4Ab06ceC7251 | ReputationRegistry — on-chain trust scores |

| Wallet | Address | Role | |--------|---------|------| | Agent / Deployer | 0x4Ba6B07626E6dF28120b04f772C4a89CC984Cc80 | Owner of ERC-8004 NFT, deployer of $XMETAV, x402 payment receiver |

---

Changelog

2026-03-08 (v28.2) — Kamino SDK Integration (Borrow/Lend + klend-sdk)

• Kamino SDK Integration — Full @kamino-finance/klend-sdk integration: SDK-first vault operations (replaced REST API), complete borrow/lending module (kamino-borrow.ts), 8 new x402 endpoints (3 free data + 5 gated lending), KaminoBorrowPanel dashboard component, OpenClaw skill files at ~/.openclaw/workspace/skills/kamino/. New deps: @kamino-finance/klend-sdk@^7.3.20, @solana/kit@^6.1.0, decimal.js@^10.6.0. @solana/kit compat via as any casts. Commit b87b67c
• Total x402 endpoints: 27 gated + 11 free = 38 (was 22 + 8 = 30)

2026-03-08 (v28/v28.1) — Cross-Chain Swaps + Kamino Vaults + Trading Dashboard

• Multi-Chain x402 Cross-Chain Swap System — Complete Base↔Solana swap pipeline: USDC on Base → CCTP bridge → Solana USDC → Jupiter Ultra swap (SOL/BONK/JUP) → Kamino vault deposit → withdraw → bridge back. 7 new modules, 6 endpoints, batch queue, job lifecycle tracking in Supabase
• Standalone Kamino Vault Endpoints — Direct POST /kamino/deposit ($0.15) and POST /kamino/withdraw ($0.15) for vault operations without full cross-chain flow
• Trading/DeFi Dashboard — New /trading page with CrossChainPanel (queue stats, swap quote tool) and KaminoPanel (vault overview, deposit/withdraw UI with Solscan links)
• Jupiter Ultra API Key — Wired API key authentication, verified: $10 USDC → 0.1225 SOL via multi-route aggregation
• DB Migration — cross_chain_jobs + cross_chain_batches tables with RLS, indexes, triggers

2026-03-03 (v26) — Scholar Research Daemon + Bridge v1.6.0

• Scholar Research Daemon — New 24/7 autonomous research agent (scholar). Cycles through 5 domains (ERC-8004, x402, Layer 2, stablecoins, SMB adoption) on scheduled intervals (15–60 min). Relevance scoring: novelty (0.35) + impact (0.30) + actionability (0.20) + recency (0.15). Auto-anchors significant findings (≥0.7) on-chain, shares high-value intel (≥0.6) to fleet memory, deduplicates against last 50 memories (70% keyword overlap). Room: Research. Color: Emerald (#10b981). Skills: deep-research, relevance-scoring, knowledge-anchoring. Arena desk at col 5, row 7. Meeting seat 75°. Connections: main, soul, oracle, midas, briefing. Runner: scripts/scholar-agent.sh. Runbook: docs/agents/scholar.md
• Bridge Daemon v1.6.0 — Upgraded from v1.5.0 with Scholar Research Loop integration, new /scholar health endpoint, graceful Scholar shutdown on SIGTERM/SIGINT
• Research Engine Modules — bridge/lib/scholar/: types (domain configs), scorer (4-dimension relevance scoring), research-loop (continuous daemon with round-robin domain scheduling), index (public exports)
• Full Stack Integration — Scholar added to: Arena (research room, desk, meeting seat, connections), office workstations, KNOWN_AGENTS, AgentChat, AgentIdentity (emerald color, 📚 room icon), ERC-8004 metadata+registration (fleet of 12, 5 new capabilities), Supabase (agent_controls + memory seeds), FLEET_AGENTS heartbeat, ALLOWED_AGENTS openclaw, both shutdown offline arrays

2026-03-03 (v25) — Sentinel Monitoring Engine + Vox Branding Agent + Bridge v1.5.0

• Vox Branding Agent — New brand voice & campaign specialist (vox). Content strategy, voice calibration, competitor monitoring. Room: OPS Outreach. Color: Cyan (#06b6d4). Skills: content-strategy, voice-calibration, competitor-analysis. Arena desk at col 1, row 5.5. Meeting seat 105°. Connections: main, midas, oracle, briefing, soul. Runner: scripts/vox-agent.sh. x402 pricing: campaign $0.15, voice audit $0.25, competitor $0.20, calendar $0.30, crisis $0.50
• Sentinel Monitoring Engine — Full autonomous monitoring system embedded in the Bridge Daemon with 6 interconnected modules:
• EventMonitor — Event-driven service health checks with adaptive polling (5s–120s) and Supabase Realtime subscriptions
• AlertManager — Anti-fatigue alerting with escalation levels (immediate → warning → critical) and configurable cooldowns
• SelfHealer — Automated remediation for downed services via launchctl kickstart, stale lock cleanup, log rotation
• PredictiveHealth — macOS resource collection (CPU, memory, disk, load), linear regression trend prediction, z-score anomaly detection
• DistributedTracer — Span-based request tracing with P95 latency, throughput, and error rate metrics
• Sentinel Orchestrator — Singleton lifecycle manager wiring all sub-systems, auto-heals on service:down events, triggers SITREP on critical alerts
• Bridge Daemon v1.5.0 — Upgraded from v1.4.0 with Sentinel integration, new /sentinel health endpoint, graceful Sentinel shutdown on SIGTERM/SIGINT
• 4 New Supabase Tables — sentinel_incidents (alert tracking), sentinel_healing_log (remediation audit), sentinel_traces (distributed spans), sentinel_resource_snapshots (system resources); all with RLS, indexes, and Realtime
• Dashboard Sentinel API — GET /api/sentinel authenticated endpoint returning health, incidents, healing log, resources, and traces
• Watchdog LaunchAgent — com.xmetav.watchdog.plist running every 5 minutes via launchd
• Dashboard CWD Fix — Fixed launchd-dashboard.sh working directory (/tmp → ${REPO}/dashboard) resolving Tailwind CSS / PostCSS build errors under launchd

2026-03-02 (v24) — Tooling + Benchmarks + Live x402 Verified

• LaunchAgent Auto-Restart — All three core services (dashboard, bridge, x402) managed by macOS launchd with KeepAlive + RunAtLoad. Auto-restart on crash with 10s throttle
• Bridge Manager Refactor — Replaced spawn() with launchctl integration for service control from dashboard UI
• Launchd Wrapper Scripts — launchd-bridge.sh, launchd-dashboard.sh, launchd-x402.sh with auto npm install, watch mode, port conflict detection
• Live x402 Payment Verification — All 4 core endpoints verified with real USDC on Base Mainnet (100% pass rate, $0.11 total)
• Service Benchmarks — Sub-10ms response times across all local services (Dashboard 7ms, Bridge 1ms, x402 1ms, Gateway 1.5ms)
• System Specs Report — Comprehensive SYSTEM_SPECS.md with hardware, services, fleet, performance metrics, and health scoring

2026-02-15 (v23) — Optimization Pass (Security + Performance + Build)

• API Security Hardening — Auth guards (requireAuth()) on all 5 data API routes (/api/soul, /api/agents/memory, /api/midas, /api/erc8004/identity, /api/anchors). UUID validation and limit clamping on user-supplied params. New shared utility src/lib/api-auth.ts with requireAuth(), isValidUUID(), clampLimit()
• RLS Policies on Dream Tables — Added authenticated SELECT policies to soul_dream_manifestations, soul_dream_sessions, soul_association_modifications (were service-role only). Added DB indexes on agent_memory(source) and memory_associations(memory_id, related_memory_id)
• Query Performance — Replaced all SELECT * with explicit column lists across dream.ts, retrieval.ts, useConsciousness.ts, dream-proposals.ts. Bounded dream cycle queries (.limit(500)). Batched N+1 association queries in dream-proposals (single query across all clusters). Batched upserts in executeManifest. getManifestationStats bounded with .limit(1000)
• Polling Reduction — useConsciousness polling interval 15s → 30s (50% reduction in dashboard Supabase load)
• Bridge Hardening — await on processPendingCommands loop (was fire-and-forget). Proper SIGTERM handler with graceful shutdown (channel unsubscribe, session teardown, PID cleanup)
• Canvas Optimizations — MemoryGraph: ctx.setTransform() replaces stacking ctx.scale() (prevents cumulative scaling bug), Map lookup for O(1) edge node resolution. DreamscapeView: document.hidden check skips rendering when tab is backgrounded
• Build Config — optimizePackageImports for lucide-react, serverExternalPackages for pg in next.config.ts. Type fix in getManifestationStats

2026-02-15 (v22) — Midas Revenue Agent + Anchor Sync + DB Consolidation

• Midas Revenue Agent — New revenue & growth intelligence agent (midas). Arena seat 165° (between Oracle 150° and Alchemist 180°). Color: Gold (#f59e0b). Skills: revenue, pricing, growth, forecast
• Revenue Dashboard — Aggregates x402 payments into daily snapshots with 7d/30d/90d forecasts, growth rates, and top-endpoint ranking
• Endpoint Analytics — Per-endpoint usage tracking: call counts, conversion rates, revenue trends (up/stable/down)
• Pricing Intelligence — Automated x402 tier analysis with confidence-scored pricing recommendations
• Growth Pipeline — ROI-ranked opportunity pipeline with category tags (endpoint/token/partnership/vertical/optimization) and status tracking
• 4 New Supabase Tables — revenue_metrics, endpoint_analytics, growth_opportunities, pricing_recommendations with RLS, indexes, triggers, Realtime
• Bridge Skill — bridge/lib/midas-revenue.ts (~330 lines): report generation, endpoint analytics, pricing analysis, growth opportunity management
• API Route — /api/midas?action=dashboard|report|endpoints|opportunities|pricing
• Agent Script — scripts/midas-agent.sh (cron-compatible: --report, --endpoints, --pricing, --growth)
• Arena Integration — Intel room tile (col 3, row 9), meeting seat, connections to oracle/alchemist/main/soul, MiniArena dot
• Anchor Sync Fix — Dashboard now keeps on-chain anchor count synced:
• AgentIdentity auto-refreshes every 30s (was mount-only)
• AnchorTimeline periodic 30s sync status refresh + re-fetches on anchor count change
• useConsciousness Supabase Realtime subscription on agent_memory INSERT where source=anchor for instant pickup
• Fixed ANCHOR_CONTRACT_ADDRESS missing from .env.local (was causing "Contract not configured")
• Switched anchors + identity API routes from cookie-based server client to admin client (service role) to bypass RLS
• DB Consolidation Migration — Formalized 5 tables + 1 view that only existed as manual setup scripts into the migration chain (agent_controls, intent_sessions, x402_payments, x402_daily_spend view, swarm_runs, swarm_tasks). All IF NOT EXISTS guarded
• Fleet Count — 12 autonomous agents + dynamic

2026-02-15 (v21) — EthSkills + Dashboard Skills UI + Build Hardening

• EthSkills Integration — 12 blockchain/Ethereum skills from ethskills.com installed across fleet agents via openclaw skills install:
• main: wallets (EOAs, Safe multisig, EIP-7702, ERC-4337, key safety)
• web3dev: tools, l2s, orchestration, addresses, concepts, security, standards, frontend-ux, frontend-playbook, building-blocks (10 skills)
• oracle: gas (L1 vs L2 tx costs), l2s (Arbitrum, Optimism, Base, zkSync, Scroll, Linea)
• midas: gas, standards, addresses, concepts (4 EthSkills for revenue context)
• Agent Identity Updates — Updated IDENTITY.md files for main (created from scratch + SOUL.md), web3dev, and oracle to include Skills sections
• ERC-8004 Metadata Update — 13 new capabilities added to on-chain metadata (ethereum-wallet-management, eip-7702-smart-eoas, safe-multisig, erc-4337-account-abstraction, l2-deployment, cross-chain-bridging, gas-economics, defi-composability, solidity-security-patterns, ethereum-standards, dapp-orchestration, frontend-ux, verified-contract-addresses, dev-tooling). Per-agent skills arrays added to fleet.agents. AgentMemoryAnchor contract added
• Dashboard Skills UI — Skills badges displayed per agent in Fleet table (purple #e879f9 badges) and Identity page fleet roster grid. New isSkill color category for skill-related capabilities
• Oracle Identity Scouting — Automated on-chain identity discovery: bridge/lib/oracle/identity-scout.ts, /api/oracle/discovery endpoint, /oracle dashboard page, useOracleDiscovery hook (10 new files)
• Arena Optimizations — Performance audit of ~3,600 line arena codebase: particle pool (object reuse), position hash diffs (skip redundant calculations), lastKnownStatus cache (prevent duplicate transitions), throttled ResizeObserver (prevent layout thrashing)
• Alchemy RPC — Switched all 11 files from default Base RPC to Alchemy (BASE_RPC_URL) for reliability
• Build Hardening — Fixed 14 pre-existing TypeScript build errors: BigInt literals (0n → BigInt(0) for ES2017 target), PromiseLike .catch(), SpeechRecognition types, Buffer/Uint8Array conversions, template literal types, duplicate object keys, unused directives. Installed missing @x402/fetch and @x402/evm dependencies. Clean npx next build passing
• OpenClaw Updated — 2026.2.12 → 2026.2.14

2026-02-14 (v20) — Cyber-Neural Memory Evolution (Crystal Materia System)

• Memory Crystal System (Materia) — Final-Fantasy-inspired memory gamification with 7 interconnected subsystems:
• Living Memory Crystals — Crystals with XP, 30 levels, star ratings (1-6★), and class evolution (anchor → mage → knight → sage → rogue → summoner → ninja → godhand)
• Crystal Fusion (FF7 Style) — 5 fusion recipes (Nexus, Prophecy, Storm, Phantom, Infinity) with animated 4-phase fusion chamber (idle → orbit → collide → birth)
• Memory Summons — Keyword-triggered crystal summoning with triple concentric summoning circle animation, converging particles, and crystal materialization
• Limit Breaks — Triggered at 10+ crystals with 500+ total XP; creates legendary 6★ godhand crystal with golden lightning banner
• Memory Cosmos World — Pannable/zoomable explorable memory landscape: golden-spiral island layout, 3 terrain types (city/wasteland/forest), neon highway bridges with data particles, crystal-topped islands, parallax stars
• Achievements — 7 seeded achievements with Bronze/Silver/Gold/Legendary tiers and progress tracking
• Daily Quests — Auto-generated daily quests with type-based objectives and XP rewards
• 6 New Supabase Tables — memory_crystals, memory_fusions, memory_summons, limit_breaks, memory_achievements, daily_quests with 3 custom enums, RLS policies, Realtime, and crystal_level_thresholds view
• Bridge Crystal Engine — bridge/lib/memory-crystal.ts (~530 lines): full game logic for XP curves, class evolution, fusion recipes, summon scoring, limit break detection, quest generation
• 7 Dashboard Components — CrystalCard (canvas-animated), CrystalInventory (filterable grid), FusionChamber (4-phase animation), SummonOverlay (summoning circle modal), LimitBreakBanner (golden lightning), MemoryCosmos (world explorer), QuestTracker (achievements + quests)
• /memory-cosmos Page — Full page assembling all crystal components with stats row, world explorer, inventory, fusion chamber, and quest tracker
• useMemoryCrystals Hook — Supabase queries + realtime subscriptions on memory_crystals and limit_breaks, 12s auto-refresh, crystal action methods
• Sidebar — Added Gem icon with Memory Cosmos nav entry

2026-02-15 (v18) — Consciousness Tab + Swap Execution + Streaming v2

• Consciousness Tab — New /consciousness route with 6-panel awareness dashboard: Unified Awareness (Main↔Soul split view), Memory Graph (force-directed canvas with drag/zoom), Anchor Timeline (on-chain anchors with BaseScan links), Context Metrics (4 metric cards + injections feed), Dream Mode Status (6hr idle threshold + insights), Mini Arena (stylized agent positions with realtime subscription). Brain icon in sidebar at position 03 (Ctrl+3). useConsciousness hook fetches from 6 Supabase tables with 15s auto-refresh
• Swap Execution System — Agent-initiated token swaps via bridge/src/swap-executor.ts. /api/swap POST endpoint. agent_swaps Supabase table for swap history. Gas balance and token balance pre-checks before submission. Clean viem error messages with failed status in chat
• Voice Swap Normalization — VOICE_ALIASES dictionary and normalizeVoiceSwap() function convert spoken swap commands ("fifty bucks of ether") to canonical form ("50 USDC for ETH")
• Streaming Pipeline v2 — 2.5× faster rendering: streamer chunk 400→160 / flush 200→80ms / first 50→30ms, executor token batching (6 tokens/15ms), useRealtimeMessages throttle 80→50ms with RAF alignment, StreamingBubble wrapped in React.memo with useMemo cleanAgentOutput, smart auto-scroll
• Identity & Schema Sync — Updated registration.json and metadata.json with swap capabilities. Supabase migrations for agent_memory and soul tables
• Memory Docs — Mermaid flowcharts in docs/memory/ (README.md, SOUL.md, ANCHORING.md)

2026-02-14 (v17) — Soul Agent + ERC-8004 Metadata Update + Arena Soul Office

• Soul Agent — New memory orchestrator agent (soul): context curation, memory association building, dream consolidation during fleet idle, and memory retrieval learning. Room: SOUL (private magenta alcove). Color: Magenta (#ff006e)
• Soul Office in Arena — Dedicated isometric alcove (cols 0–1, rows 2–5) with magenta floor tiles, glass partition walls, "SOUL" room label, L-shaped surveillance desk, and an arc of mini fleet-monitor screens that mirror every agent's state in real-time
• Soul Arena Integration — Soul agent node, meeting seat (observer position at 195°), and topology connections to main, briefing, oracle, alchemist, sentinel (watches everyone)
• ERC-8004 Metadata Update — Added Soul to fleet.agents and fleet.rooms in on-chain metadata. Added 5 soul-specific capabilities: soul-memory-orchestration, dream-consolidation, memory-association-building, context-packet-curation, memory-retrieval-learning
• Soul DB Schema — Three Supabase tables (memory_associations, memory_queries, dream_insights) with RLS policies for Soul's memory orchestration layer
• Soul Bridge Library — Full bridge/lib/soul/ module: context building, memory retrieval, association building, dream mode (idle consolidation), and type definitions
• Sidebar Cleanup — Removed Soul tab from sidebar navigation (no /soul page yet; Soul is an arena-only agent)
• Fleet Count — 11 autonomous agents (main, sentinel, soul, briefing, oracle, alchemist, web3dev, akua, akua_web, basedintern, basedintern_web) + dynamic

2026-02-13 (v15) — Sentinel Agent + Agent Memory + Identity System + Noise Filter

• Sentinel Agent — New fleet lifecycle manager (sentinel): spawn coordination, resource management, inter-agent communication, fleet health monitoring. Room: COMMAND. Color: Red (#ef4444). Commands: status, health, spawn, queue, errors
• Agent Identity System — Created IDENTITY.md + SOUL.md for all sub-agents (web3dev, oracle, briefing, alchemist, sentinel). Agents now have full self-awareness: who they are, what commands they have, their team, operating principles, and communication style
• Main Agent Memory — Persistent daily session (dash_main_YYYYMMDD) gives main conversation context across commands within a day. Falls back to unique session ID when lock is held (concurrent commands don't deadlock)
• Noise Filter v3 — Added [diagnostic], [heartbeat], [bridge], [swarm], [intent-tracker], [voice/...], and session-lock errors to output filter. Agent responses now show only actual content
• Voice STT Overhaul — Chrome's built-in Web Speech API (SpeechRecognition) as default STT. Whisper-1 fallback with language: "en". Removed prompt/temperature that caused hallucination loops
• Operator Cleanup — Removed operator from agent tables, spawn list, and ERC-8004 registration (operator is the human user, not an agent)
• Web3 Lab — Dedicated cubicle for web3dev with orange floor tint and glass partitions (cols 7-9, rows 2-