Open-source first AI inference — own your inference forever via the Morpheus decentralized network.
Canonical repository: profbernardoj/morpheus-skill
EverClaw (Morpheus Skill) connects your OpenClaw agent to the Morpheus decentralized inference network — putting open-source models like GLM-5 front and center as your default, with Claude as a fallback only when needed.
Your agent runs on inference you own: GLM-5, GLM-4.7 Flash, Kimi K2.5, and 30+ models powered by staked MOR tokens that recycle back to you.
This is a monorepo that produces 28+ flavor repos via composition:
packages/
core/ # Common Morpheus infrastructure
scripts/ # All shared scripts (proxy, session, wallet, security, setup)
tests/ # All tests
references/ # API docs, economics, models
docs/ # Documentation site
templates/ # Boot templates, system configs, flavor persona templates
config/ # Default OpenClaw configs
SKILL.md # Core skill definition
Dockerfile # Container build
docker-compose.yml # Docker compose
everclaw-docker/ # Docker-specific build configs
everclaw-key-api/ # Vercel key API service
flavors/
morpheus-skill/ # The canonical/default flavor
androidclaw.org/ # Android ecosystem flavor
bitcoinclaw.ai/ # Bitcoin ecosystem flavor
emailclaw.org/ # Email management flavor
ethereumclaw.com/ # Ethereum ecosystem flavor
... # 28 total flavor directories (one per domain)
scripts/
ecosystem-sync.sh # Composes core + flavor → pushes to flavor remotes
flavor-compose.sh # Composes a single flavor into a deployable repo
skills/ # Bundled skills (security, chat, prompt-guard, etc.)
archive/ # Archived: alternative installers, marketing, one-time tools
packages/core/contains all shared infrastructure: scripts, tests, docs, templatesflavors/<name>/contains flavor-specific files: README.md, flavor.json, persona templatesscripts/ecosystem-sync.shcomposes each flavor by merging core + flavor → pushes to that flavor's remote- Canonical remotes (
origin,everclaw-org) receive the full monorepo - Flavor remotes receive only the composed output (core + their specific flavor)
curl -fsSL https://get.everclaw.xyz | bashgit clone https://github.com/profbernardoj/morpheus-skill.git
cd morpheus-skill
npm install
node packages/core/scripts/setup.mjsEach flavor is a standalone repo with its own domain and persona:
| Flavor | Domain | Description |
|---|---|---|
| morpheus-skill | morpheusclaw.com | Canonical default |
| bitcoinclaw.ai | bitcoinclaw.ai | Bitcoin ecosystem |
| ethereumclaw.com | ethereumclaw.com | Ethereum ecosystem |
| glmclaw.com | glmclaw.com | GLM model focus |
| emailclaw.org | emailclaw.org | Email management |
| ... | ... | 28 total flavors |
See flavors/ for the complete list with flavor.json configs.
# Dry run — see what would happen
./scripts/ecosystem-sync.sh --dry-run
# Push to all remotes
./scripts/ecosystem-sync.sh
# Push only one flavor
./scripts/ecosystem-sync.sh --flavor bitcoinclaw.ai./scripts/flavor-compose.sh flavors/bitcoinclaw.ai /tmp/composed/bitcoinclaw.ainpm test- Create the flavor directory:
mkdir -p flavors/my-new-flavor
- Create
flavors/my-new-flavor/flavor.json:{ "name": "My New Flavor", "slug": "my-new-flavor", "domain": "my-new-flavor.com", "description": "What this flavor does", "remote": "https://github.com/org/my-new-flavor.git", "defaultModel": "glm-5", "persona": "Short persona description" } - Create
flavors/my-new-flavor/README.mdwith flavor-specific documentation. - (Optional) Add
flavors/my-new-flavor/templates/with custom SOUL.md, IDENTITY.md, etc. - Add a git remote:
git remote add my-new-flavor https://github.com/org/my-new-flavor.git - Sync:
./scripts/ecosystem-sync.sh --flavor my-new-flavor
- Morpheus Proxy — OpenAI-compatible proxy with auto-session management
- Model Router — 3-tier routing (light/standard/heavy), open-source first
- Wallet Management — Zero-dependency via macOS Keychain
- Gateway Guardian — Health watchdog with billing-aware escalation
- Security — PII guard, gateway guardian, bundled security skills
- x402 Payments — Agent-to-agent USDC payments
- ERC-8004 Registry — Discover trustless agents on Base
- Local Fallback — Hardware-aware Ollama with auto model selection
- Buddy Bots — Multi-agent coordination and provisioning
- Three-Shift Engine — Cyclic task execution (6 AM / 2 PM / 10 PM)
MIT