WAKE Protocol

WAKE MCP Server

[npm] [License]

Will-Aware Knowledge Execution - A post-mortem protocol for AI agents.

23 tools · SQLite + AES-256-GCM encryption · Token auth

Quick Start

# From sourcepnpm install && pnpm run buildpnpm start              # stdio + dashboard on :3000pnpm run start:http     # dashboard only (no stdio)pnpm test               # 69 tests
# npx (no install needed)npx wake-mcp-servernpx wake-mcp-server --http
# Global installnpm i -g wake-mcp-serverwake-mcp-serverwake-mcp-server --http
# Dockerdocker build -t wake-mcp-server .docker run -p 3000:3000 -v wake-data:/app/data wake-mcp-server

Connect Your Agent

Warp / Claude Desktop / Cursor — same config, different file locations:

Using npx (recommended):

{  "wake-protocol": {    "command": "npx",    "args": ["wake-mcp-server"]  }}

Or with a local build:

{  "wake-protocol": {    "command": "node",    "args": ["<path>/wake-mcp-server/build/index.js"]  }}

Any MCP client over HTTP:

http://localhost:<port>/mcp

Default port is 3000. If taken, auto-increments. Set WAKE_PORT to override. Check the console output for the actual URL.

Agent Skill

SKILL.md — platform-agnostic instructions that teach any agent how to use WAKE:

• Owner — auto-heartbeat, conversational will setup, proactive knowledge contribution
• Verifier — death confirmation handling
• Beneficiary — tier-scoped access, final messages, handoff packages
• Executor — EULOGY management, legal export, terminal state, data purge

Point your agent at SKILL.md or include it in your system prompt.

Protocol

ACTIVE → VIGIL → EULOGY → REST

• ACTIVE — Owner alive. Agent calls heartbeat each interaction.
• VIGIL — Inactivity threshold exceeded. Awaiting verifier confirmation.
• EULOGY — Succession executing. Tiered access, Black Box, final messages.
• REST — Terminal state (archive/distill/delete). Protocol complete.

Authentication

Tokens generated once during configure_will. Three types:

• Master — owner operations (heartbeat, configure, vigil)
• Verifier — death confirmation only
• Beneficiary — per-person, scoped by tier (executor/beneficiary/memorial)

Tokens stored as SHA-256 hashes. State encrypted with AES-256-GCM (server key).

Tools

Will Management

Protocol Progression

Data Access

Knowledge

Inter-Agent Handoff

Backup & Recovery

Legal & Compliance

Key Features

• Multi-user — ownerId param on every tool, independent state per owner
• Auto-backup — previous state saved on every write, last 5 kept
• Background monitor — checks heartbeat staleness every 5 min, auto-triggers VIGIL
• Dead man's switch — auto-escalates VIGIL → EULOGY if verifier doesn't respond
• Webhooks — fire on vigil.triggered, eulogy.started, eulogy.message, rest.executed, timelock.released
• Time-locks — releaseAfter on messages and knowledge entries, sealed until date
• Redaction filtering — entries matching redaction categories hidden from all tiers
• Tier-scoped Black Box — executor sees all, beneficiary sees finances/contacts/documents, memorial sees curated entries only
• No-resurrection directive — enforceable flag propagated to handoffs and certificates
• Jurisdiction mapping — legal export maps to applicable laws (RUFADAA, GDPR, ELVIS Act, etc.)
• Deletion certificate — SHA-256 attested record, sole survivor of data purge

Dashboard

Web UI at http://localhost:3000. Role-based views:

• Login — paste any token, auto-detects role
• Owner — status, knowledge CRUD, heartbeat, VIGIL trigger, legal export
• Verifier — status, death confirmation
• Executor — Black Box, messages, audit, terminal state, handoff, purge
• Beneficiary/Memorial — tier-scoped Black Box, messages

Set WAKE_PORT=8080 to change the port.

Data

SQLite database at data/wake.db. Server encryption key at data/.server-key. Deletion certificates at data/deletion-certificate-*.json.

Delete the data/ directory to reset everything.

Deploy

One-click:

[Deploy on Railway]

[Deploy on Fly.io]

CLI:

# Railwayrailway init && railway up
# Fly.iofly launch       # creates app + volumefly deploy
# Dockerdocker build -t wake-mcp-server .docker run -p 3000:3000 -v wake-data:/app/data wake-mcp-server

Docker Compose:

services:  wake:    build: .    ports: ["3000:3000"]    volumes: ["wake-data:/app/data"]volumes:  wake-data:

Config files included: railway.toml, fly.toml, vercel.json, Dockerfile.

Note: Vercel is serverless — SQLite won't persist. Use Railway or Fly for production.