⚠️ Everclaw is experimental beta open source software for testers only. Use at your own risk.

Everclaw

v0.9.7

AI inference you own, forever powering your OpenClaw agents via the Morpheus decentralized network.

Get Started

Choose your starting point:

🆕

New to OpenClaw

Install OpenClaw first, then add Everclaw

Already have OpenClaw

Add Everclaw to your existing setup

Install OpenClaw + Everclaw

1

Install OpenClaw

OpenClaw is the AI agent framework that Everclaw extends. Install it with npm:

npm install -g openclaw
2

Run the OpenClaw Setup Wizard

This configures your agent, sets up your model provider, and creates your workspace:

openclaw onboard
3

Install Everclaw

Install from ClawHub (easiest) or the command-line installer:

clawhub install everclaw-inference

Or: curl -fsSL https://raw.githubusercontent.com/profbernardoj/everclaw/main/scripts/install-everclaw.sh | bash

4

Install the Morpheus Router

Downloads the Morpheus proxy-router for your platform:

bash ~/.openclaw/workspace/skills/everclaw/scripts/install.sh
5

Set Up the Proxy & Guardian

Installs the OpenAI-compatible proxy and the gateway health monitor:

bash ~/.openclaw/workspace/skills/everclaw/scripts/install-proxy.sh
6

Create Your Wallet

Generates an Ethereum wallet and stores the private key in your macOS Keychain. No external accounts needed.

node ~/.openclaw/workspace/skills/everclaw/scripts/everclaw-wallet.mjs setup
7

Fund Your Wallet

Send ETH to your wallet address (shown after setup). You need a small amount of ETH on Base for gas and to swap for MOR tokens. Then swap:

# Swap 0.05 ETH for MOR tokens node ~/.openclaw/workspace/skills/everclaw/scripts/everclaw-wallet.mjs swap eth 0.05 # Approve MOR for staking node ~/.openclaw/workspace/skills/everclaw/scripts/everclaw-wallet.mjs approve
8

Start Inference

Launch the Morpheus router. Your agent now has access to decentralized AI inference.

bash ~/.openclaw/workspace/skills/everclaw/scripts/start.sh

💡 MOR is staked, not spent

When you open an inference session, your MOR tokens are locked as a deposit. When the session ends, they come back. Stake them again for a new session. This is what makes the access yours forever — you're not buying consumable API credits.

Add Everclaw to OpenClaw

1

Install Everclaw

Install from ClawHub or the command-line installer:

clawhub install everclaw-inference

Or: curl -fsSL https://raw.githubusercontent.com/profbernardoj/everclaw/main/scripts/install-everclaw.sh | bash

2

Install Router, Proxy & Guardian

Two commands set up everything — the Morpheus router, OpenAI-compatible proxy, and gateway health monitor:

bash ~/.openclaw/workspace/skills/everclaw/scripts/install.sh bash ~/.openclaw/workspace/skills/everclaw/scripts/install-proxy.sh
3

Create Your Wallet

Generates a wallet and stores the key in macOS Keychain. No 1Password, no external accounts.

node ~/.openclaw/workspace/skills/everclaw/scripts/everclaw-wallet.mjs setup
4

Fund & Swap

Send ETH (on Base) to your wallet address, then swap for MOR:

# Swap ETH for MOR node ~/.openclaw/workspace/skills/everclaw/scripts/everclaw-wallet.mjs swap eth 0.05 # Approve MOR for staking node ~/.openclaw/workspace/skills/everclaw/scripts/everclaw-wallet.mjs approve
5

Start & Configure Fallback

Launch the router, then tell your agent to use Morpheus as a fallback model:

# Start the Morpheus router bash ~/.openclaw/workspace/skills/everclaw/scripts/start.sh # Check it's working curl http://127.0.0.1:8083/health

Then add Morpheus as a fallback in your openclaw.json — see the full configuration guide.

💡 MOR is staked, not spent

When you open an inference session, your MOR tokens are locked as a deposit. When the session ends, they come back. Stake them again for a new session. This is what makes the access yours forever — you're not buying consumable API credits.

How It Works

Stake MOR tokens for inference sessions. When the session ends, your MOR comes back. Stake again. Forever.

Your Agent
Everclaw Proxy
Morpheus P2P
AI Model

🆕 What's New in v0.9.7

🔍 Diagnostic Tool (diagnose.sh)

One command, 18 checks, every known failure mode. Config validation catches bad provider prefixes, missing Morpheus providers, and the reasoning: true bug. Infrastructure checks verify the proxy-router, proxy, blockchain sessions, MOR balance, API Gateway, launchd services, and live end-to-end inference. Run bash scripts/diagnose.sh to know exactly what's broken and how to fix it.

🛡️ Provider Misconfiguration Guard (v0.9.6)

Three layers of defense against the #1 onboarding issue: using everclaw/ as a model prefix. The bootstrap script auto-fixes it, both install scripts validate post-install, and diagnose.sh catches it at any time. Always-on proxy-router (v0.9.5) with launchd KeepAlive ensures Morpheus fallback never silently dies.

v0.9.4 — Session Archiver

Size-triggered archiving + Guardian v4

v0.9.1 — Multi-Key Auth

API key rotation + circuit breaker

v0.8 — Gateway Bootstrap

Zero-config instant inference startup

v0.7 — Agent Economy

x402 payments + ERC-8004 registry

v0.6 — Model Router

3-tier classifier, 13 dimensions, <1ms

v0.5 — Proxy Resilience

Cooldown-safe errors, session auto-retry

v0.4 — Zero-Dep Wallet

macOS Keychain, no external accounts

v0.3 — Security Suite

SkillGuard, ClawdStrike, PromptGuard, Bagman

What's Included

🚀

Zero-Config Bootstrap

One command gives new users instant AI inference via the Morpheus API Gateway. No API key, no wallet, no node — just run and go.

🔑

Multi-Key Rotation

Add multiple Venice API keys. When one runs out of DIEM credits, auto-rotates to the next. Stay on Claude instead of falling to cheaper models.

♾️

Decentralized Inference

Access Kimi K2.5, GLM-4.7 Flash, Qwen3, and 30+ models through the Morpheus P2P network.

🔌

OpenAI-Compatible Proxy

Standard API format. Auto-opens 7-day sessions, renews before expiry, auto-retries on session errors, and returns cooldown-safe error codes.

🧭

Model Router

13-dimension prompt classifier runs in <1ms. Routes cron jobs to GLM Flash (owned), research to Kimi K2.5 (owned), complex reasoning to Claude (premium).

💸

x402 Payments

Pay any x402-enabled agent with USDC on Base. Automatic 402 detection, EIP-712 signing, budget controls, and dry-run mode.

📋

ERC-8004 Registry

Discover agents on-chain. Read identity, services, endpoints, and reputation from the Base mainnet registries.

🛡️

Gateway Guardian v4

Billing-aware escalation — classifies errors, skips useless restarts for credit exhaustion. Through-OpenClaw probes, credit monitoring, circuit breaker, fixed restart chain. Signal notifications with DIEM reset ETA.

🔐

Zero-Dependency Wallet

Private keys stored in macOS Keychain. No 1Password, no external accounts. Generate, swap, and stake in one script.

📦

Smart Session Archiver

Size-triggered archiving prevents dashboard overload. Monitors session size, archives old sessions above 10MB. Runs as cron — no-op when healthy.

🔍

SkillGuard Scanner

Scan skills for malicious patterns before installing. Protects against supply chain attacks.

🧱

Prompt Injection Defense

Multi-language detection, severity scoring, and HiveFence threat intelligence network.

Available Models

Model Type Notes
Kimi K2.5GeneralRecommended — most reliable
Kimi K2 ThinkingReasoningExtended thinking for complex tasks
GLM-5GeneralZhipu's latest flagship 🆕
GLM-5:webGeneral + WebGLM-5 with web search 🆕
MiniMax M2.5GeneralMiniMax's latest model 🆕
MiniMax M2.5:webGeneral + WebM2.5 with web search 🆕
GLM 4.7 FlashFastLower latency
GLM 4.7GeneralFull GLM model
Qwen3 235BGeneralLarge parameter count
GPT-OSS 120BGeneralOpenAI's open-source model