The sovereign intelligence substrate — persistent memory, identity, and attested governance for your whole AI fleet.
One brain and one rulebook shared by every coding agent you run: Claude Code, Cursor, Codex, Gemini CLI, OpenCode, Antigravity. Per-tool memory is now table stakes; what your fleet is missing is cross-tool, cross-repo, governed, and attested. That's this repo — 144 agents, 83 auto-activating skills, 6 semantic vaults, an MCP server, and a provenance protocol (SIP). 965 tests keep the claims honest.
Note
The Ultimate Substrate for Sovereign Agent Armies & Web4 Intelligence Fabrics. SIS is the open protocol (SIP) + reference build for building, governing, and operating production-grade "Mind + Mesh + Steward" agent armies — persistent, attested, Genius-grounded swarms across every surface. The unclaimed category: the operator's full sovereign intelligence estate (not rented vendor agents or brittle glue frameworks).
Ethereum for intelligence: Open SIP protocol for adoption flywheel + layered value capture (protected canon/encoded-self + commercial Estate Factory delivery). Trinity is instance #1. Factory economics: 80% reuse after first estate.
Exposing: 144 agents, 83 auto-activating skills, 6 semantic vaults, MCP server, provenance (SIP), and now the repeatable Estate Factory (4-layer blueprint, starlight-estate-os profile, /estate-blueprint + /estate-steward commands, full E2E commissioning workflow).
See docs/strategic/sip-web4-substrate-strategy.md (IP/Web4 model), docs/delivery/estate-army-commissioning-workflow.md (full client workflow + benefits vs DIY/competitors), templates/estate-os/ (thin reusable profile), docs/boards/2026-06-16-estate-factory-web4-positioning-verdict.md (PROCEED-WITH-REVISE, REVISE items closed).
The pull: "Help me set up and run a real, reliable, compounding agent army that acts as an extension of me — with memory that lasts, rules I control, production reliability." SIS + your depth delivers it E2E at world-class standards.
# 1. Seed the six JSONL vaults (~/.starlight/vaults)
npx -p @arcanea/starlight-intelligence-system starlight init --vaults// 2. Point any MCP client at them (Claude Code, Cursor, Codex, ...)
{
"mcpServers": {
"starlight": {
"command": "node",
"args": [
"node_modules/@arcanea/starlight-intelligence-system/dist/mcp-server.js",
"--vault-dir", "~/.starlight/vaults"
]
}
}
}Restart your client: ten sis_* tools in every session — the same vaults, from every CLI you run. Full walkthrough below in Quick start; measured receipts in BENCHMARKS.md.
| Layer | What it is | What lives here | License | Adopt how |
|---|---|---|---|---|
| Substrate (SIP) | A six-layer protocol that lets sovereign parties compose intelligence systems without losing sovereignty. | SIP.md, SIS.md, ALLIANCE.md, STACK.md, VOICES.md, VERTICALS.md, MEMORY.md, REGISTRY.md, SKILL.md, .claude/commands/ |
MIT | Read SIP.md, attest with /sip-attest, fork what you need. |
| Operational (reference build) | This repo's working implementation: 144 named agents (7-archetype council + specialist tiers + Hermes search + evaluator + Energy Intelligence + Social Intelligence), 6 semantic vaults, 83 auto-activating skills across 16 domains, MCP server, 14 strategic commands + 100+ slash commands, seven-platform adapters. Frank's daily-driver. | agents/, memory/, skills/, commands/, core/, context/, src/ (npm package) |
MIT | Install @arcanea/starlight-intelligence-system, run the MCP server, write to your vaults. |
You can adopt just the substrate (fork SIP for your own work), just the operational layer (use the MCP server for AI memory), or the full stack (Frank's reference build, end to end). They are independent.
Operator? Start at SETUP.md — covers
private/instance state, Infisical Path A vs env-var Path B, Windows + Linux cron wiring, cockpit launch, and a smoke test, end-to-end in roughly 30 min.New to the protocol? Don't fork this repo. Fork the SIP adoption kit — eleven markdown files, no code, ship your first attested artifact in 60 seconds. Compose upward when you're ready.
New in v8.3.0 (2026-06-12): First-run experience hardened so the package works for anyone who installs it.
starlight init --vaultsseeds the six JSONL vaults (and the MCP server auto-seeds an empty--vault-diron first boot), so a fresh install is never silently empty.sis_searchis now honestly described as keyword + temporal (no embeddings); a measured retrieval recall@k harness (npm run eval:retrieval) grounds the bm25 claim in CI; and an optionalsqlite-vecsemantic layer is roadmapped indocs/bring-your-own-model.md. SeeCHANGELOG.md § v8.3.0andRELEASING.md.New in v8.1.0 (2026-05-17): Composition Layer primitive declared in
STACK.md— universal IS may compose over its Domain Sub-Stacks via commands at the IS-itself. Wealth IS v0.2 evolved as first composition-layer reference. Crypto Intelligence v0.1 shipped as third reference Domain Sub-Stack (after People + Sound) with Houses-as-sub-systems primitive — House of On-Chain live with 5 commands./blessglobal skill + chronicle infrastructure initialized. 10-IS taxonomy invariant preserved. SeeCHANGELOG.md § v8.1.0+docs/boards/2026-05-17-crypto-investment-spawn.md.
2026-06-12 Grok Queen visual + continuous advance (whole SIS): First-class surfaces (
/starlight-queen/sq /so /starlight architect/queen) + executable driver making ROUTE→MEASURE→LEARN→RATIFY→LEDGER live. 5 parallel premium visuals (Queen loop+gateway+3D MemPalace, SIS arch, routing heatmap, advance receipt). Routing evo (Grok classes + memory-consolidation-queen + palace-visual-recall). /si now visual surface. Memory deeper (Queen drives consolidation + visual recall + gateway state). Surgical updates across vaults, HARNESS, agents, VAULT_ARCH, docs. This advance Queen-driven end-to-end. Drive withnpm run queen .... All Built on SIP. See operational-vault.md [2026-06-12 Queen Advance], tools/queen/queen-advance-2026-06-12.json, commands/starlight-queen.md.
flowchart TB
Human["Human intent"]
Vaults["6 semantic vaults<br/>JSONL source of truth"]
Index["SQLite + FTS5<br/>rebuildable index"]
MCP["MCP server<br/>10 sis_* tools"]
Adapters["Claude Code · Cursor · Codex · Gemini · OpenCode"]
Repos["Repo-specific agent harnesses"]
Output["Safer, memory-aware agent work"]
Human --> Vaults
Vaults --> Index
Index --> MCP
MCP --> Adapters
Adapters --> Repos
Repos --> Output
Output --> Vaults
| Capability | What It Gives Agents |
|---|---|
| Semantic vaults | Durable memory across sessions and tools |
| Temporal confidence | Old knowledge decays unless reconfirmed |
| Contradiction detection | Conflicting memories become visible |
| Platform adapters | Same substrate across Claude, Cursor, Codex, Gemini, OpenCode |
| MCP server | Tool-native access to memory and retrieval |
| Harness checks | Prompt surfaces stay aligned with reality |
SIS + the new Estate Factory (post 2026-06-16 Board) is the repeatable system to commission and run full "Mind + Mesh + Steward" sovereign intelligence estates / agent armies for principals, alliances, communities.
- Mind: Grounded 10-IS (Genius excavation, Second Brain, Brand, Wealth DPI, Creator pipelines, etc.) + Orchestrator as router.
- Mesh: Production swarms — 6 orchestration patterns, /si multi-CLI routing (Claude/Codex/Gemini/Antigravity/Grok...), Hermes retrieval, council/Prime synthesis, claws for ops/attestation/sentinel, yolo/hive for autonomy, amplification Claws with voice-lock/frequency caps.
- Steward: Ongoing ops, health (evals/drift), evolution, board facilitation. Retainer annuity.
4-Layer Blueprint (Persona/Naming, Topology/Swarm shape, Kernel/insight density, Modules/verticals via spawn-domain-stack) + Genius + Freedom Path grounding = the "thin tuned 20%" that makes it yours, not generic.
Reusable 80%: templates/estate-os/ profile (file contract, 10-IS, orchestrator harnesses, claws, Memory Bus + Veil, attestation, naming skins, module scaffolds). Fork per client as estate-.
Commands: /estate-blueprint (4-layer config + build brief), /estate-steward (setup/health/evolve/report for Standing phase), /estate-army-deploy (production deploy, Steward runtime harness, scale hooks — see evolutions).
Full E2E + Factory evolutions: docs/strategic/estate-factory-evolutions.md (post-REVISE track). See PR #22 for the complete delivery.
Full E2E: Front door → Excavation → Blueprint → SOW (with alliance vs commercial split) → Scaffold (sovereign-spawn or estate-os) → /si-routed build + Pilot → Scale → Handover + Steward retainer.
Benefits vs DIY/competitors: True sovereignty + encoded-self protection (SIP §5.7), human-grounded (not commodity), full life stack + domain depth (proven verticals: People/Sound/Music IS/Crypto/Energy), production substrate (not toys), compounding via "Built on SIP" + promotion loop, E2E process with gates/boards at highest standards. Factory: estate #2 fraction of effort.
Trinity: Instance #1 (alliance governance abundant + separate commercial SOW for advanced tech/army). See docs/delivery/trinity-management-playbook.md, docs/delivery/estate-sow-template.md.
See full: docs/strategic/sip-web4-substrate-strategy.md, docs/delivery/estate-army-commissioning-workflow.md, upgrades track, hero plan.
Run the local harness guard:
npm run agents:harness-checkStarlight Intelligence Protocol — the contract that lets sovereign parties compose intelligence systems without losing sovereignty. Six layers:
- File contract — canonical names and shapes for
SKILL.md,AGENTS.md/VOICES.md,MEMORY.md,CANON.md,SOUL.md,STACK.md,.claude/commands/*, plus.arc/.nea/.skillextensions. - Attestation protocol — verifiable "Built on SIP" attribution via
/sip-attest. Refuses decorative use. - MCP registry standard — how MCP servers declare, compose, version (
mcp.jsonschema). - Command taxonomy — protocol / alliance / vertical / sovereign tiers with explicit decision rights.
- Sovereignty + attribution clause — the non-negotiable social contract.
- Archetype extension — optional canon adoption (Arcanea archetypes available CC-BY-NC).
Canonical URL: starlightintelligence.org/protocol · Source of truth: SIP.md in this repo.
| File | What it does |
|---|---|
SIP.md |
The protocol spec (six layers). |
SIS.md |
Substrate map — verticals, composition rules, what SIS does and doesn't provide. |
ALLIANCE.md |
The alliance forging method — how 2-5 sovereign nodes compose without collapsing into one entity. |
STACK.md |
Recommended sovereign stack (L0-L6). Defaults, not mandates. |
VOICES.md |
Five canonical voice archetypes (architect, sovereign-creator, protocol-defender, implementer, overseer). |
VERTICALS.md |
Public registry of sovereign verticals + alliance class definitions. |
MEMORY.md |
Template for per-instance state. |
REGISTRY.md |
MCP server registry. |
SKILL.md |
Substrate-layer behavior (what AI adopts when working at this layer). |
.claude/commands/ |
9 reference slash commands (/sip-attest, /alliance-forge, /alliance-reflect, /alliance-decide, /vertical-spawn, /luminor-board, /sovereign-signal, /openclaw-audit, /wealth-dpi). |
# Pressure-test before committing
/luminor-board "Open-source the agent layer or keep it closed?"
# Forge an alliance under SIP
/alliance-forge trinity "architect,sovereign-creator,protocol-defender,implementer"
# Spawn a vertical IS under SIS
/vertical-spawn anime-legends "anime-aesthetic fiction + character design"
# Attest a shipped artifact
/sip-attest path/to/artifact.mdEvery cross-party artifact carries the "Built on SIP" attestation block. Silent composition is a breach.
This repo also ships Frank's working implementation — the daily-driver intelligence system that runs on top of SIP. Use it directly, fork it, or replace it entirely with your own layer above the substrate.
Option 1: As an MCP server (recommended for AI tools)
First, seed your vault directory so the system has somewhere to read from
(a fresh install has no ~/.starlight/vaults yet):
npx -p @arcanea/starlight-intelligence-system starlight init --vaults
# → seeds the six JSONL vaults in ~/.starlight/vaults (welcome entry + public examples)Then point your MCP client at that directory:
{
"mcpServers": {
"starlight": {
"command": "node",
"args": [
"node_modules/@arcanea/starlight-intelligence-system/dist/mcp-server.js",
"--vault-dir",
"~/.starlight/vaults"
]
}
}
}Restart Claude Code. You now have ten sis_* tools available in every session.
(If you skip the seed step the server still works — it auto-seeds an empty
--vault-dir on first boot so the empty state is never silently broken.)
Option 2: As a library
npm install @arcanea/starlight-intelligence-systemimport { StarlightIntelligence } from "@arcanea/starlight-intelligence-system";
import { createAdapter } from "@arcanea/starlight-intelligence-system/adapters";
const sis = new StarlightIntelligence();
sis.initialize();
sis.remember({
content: "Always Read a file before editing — catches stale state",
category: "pattern",
tags: ["workflow", "edit-safety"],
confidence: 0.95,
});
const adapter = createAdapter("claude-code");
const context = await adapter.generate({ vaultDir: "~/.starlight/vaults" });| Vault | Symbol | Purpose | Example entry |
|---|---|---|---|
| Strategic | ◆ | Business insights, architecture decisions, competitive moats | "Open Core + Founding Circle beats premium tiers at this stage" |
| Technical | ⬡ | Implementation learnings, stack decisions, patterns | "SQLite FTS5 with bm25 beats embeddings for <10k entries" |
| Creative | ✦ | Design preferences, aesthetic rules, voice, lore | "Never Cinzel. Space Grotesk display, Inter body." |
| Operational | ▸ | Workflow patterns, execution lessons, process rules | "Max 2 worktrees. Digest pattern for terminal output." |
| Wisdom | ◎ | Deep principles, truths, cross-domain insights | "Memory that compounds is intelligence that grows" |
| Horizon | ↗ | Vision statements, append-only ledger of human intentions | "Build the substrate that makes AI agents continuous" |
Each vault is a JSONL file. Human-readable. Git-versionable. Greppable.
Where memory lives. The MCP server and
src/retrieval.tsread*.jsonlfiles in your--vault-dir(default~/.starlight/vaults) — that is the one source of truth at runtime. The repo ships two example sets for reference only:public-vault/*.jsonlis the canonical starter content thatstarlight init --vaultscopies into your vault dir;memory/vaults/*.mdare human-readable narrative snapshots and are not read by the engine. Seed once, then your own~/.starlight/vaultsis what counts.
- SQLite hybrid retrieval —
src/retrieval.tsbuilds a rebuildable FTS5 shadow index over JSONL vaults with bm25 ranking. Keyword + temporal today (measured recall in CI vianpm run eval:retrieval); optionalsqlite-vecsemantic layer is roadmapped indocs/bring-your-own-model.md. - Temporal reasoning —
src/temporal.tsadds validity windows and a 90-day confidence half-life. - Contradiction detection —
src/contradiction.tsfinds conflicting entries via word-trigram Jaccard with opposing-signal boosting. - Dreaming —
src/dreaming.tsprocesses session transcripts in the background. - Five platform adapters — Claude Code, Cursor, Codex, Gemini CLI, OpenCode share the same vaults.
- MCP v2 —
src/mcp-server.tsships ten tools over JSON-RPC 2.0 stdio.
| Tool | Description |
|---|---|
sis_vault_search |
Free-text search across vaults |
sis_recent_entries |
Latest entries from one or all vaults |
sis_stats |
Total entry counts per vault |
sis_append_entry |
Write a new entry to a vault |
sis_entry_types |
List supported vault types and entry categories |
sis_search |
Keyword + temporal search (term-overlap score, tag boost, staleness penalty) |
sis_confirm |
Touch lastConfirmed on an entry |
sis_invalidate |
Mark an entry as expired |
sis_contradict |
Flag two entries as contradictory |
sis_stale |
List entries not confirmed within a threshold |
| Platform | Memory file | MCP config | Max tokens |
|---|---|---|---|
| Claude Code | CLAUDE.md |
~/.claude/settings.json → mcpServers |
200,000 |
| Cursor | .cursorrules |
Cursor settings → MCP | 128,000 |
| Codex | AGENTS.md |
~/.codex/config.toml |
192,000 |
| Gemini CLI | GEMINI.md |
~/.gemini/settings.json |
1,000,000 |
| OpenCode | AGENTS.md (compact) |
~/.opencode/config.json |
128,000 |
The reference build maps SIP's 5 voice archetypes to 7 named runtime agents — Orchestrator, Prime, Architect, Navigator, Sentinel, Weaver, Sage. Full registry: agents/AGENT_REGISTRY.md.
Voice archetypes are abstract; named agents are specific implementations. Anyone forking SIP can choose entirely different agents above the substrate.
| If you want to... | Take | Leave |
|---|---|---|
| Adopt SIP for your own creator stack | SIP.md, SKILL.md, VOICES.md, .claude/commands/sip-attest.md, .claude/commands/alliance-forge.md |
Everything in agents/, memory/, skills/, src/ |
| Use the MCP memory server, no protocol | src/, dist/, npm package |
Substrate spec files |
| Forge an alliance under SIP | SIP.md, ALLIANCE.md, VOICES.md, all 9 commands in .claude/commands/ |
Operational layer |
| Run the full reference build | All of it | Nothing |
starlightintelligence.org/protocol mirrors SIP.md. This repo is the source of truth.
┌─────────────────────────────────────────┐
│ JSONL vaults (source of truth) │
│ ~/.starlight/vaults/*.jsonl │
│ human-readable · git-versionable │
└────────────────┬────────────────────────┘
│
│ rebuildable from JSONL
▼
┌─────────────────────────────────────────┐
│ SQLite + FTS5 (shadow index) │
│ bm25 ranking · temporal filters │
└────────────────┬────────────────────────┘
│
│ JSON-RPC 2.0 over stdio
▼
┌─────────────────────────────────────────┐
│ MCP server (10 sis_* tools) │
└────────────────┬────────────────────────┘
│
┌────────────────────┼────────────────────────┐
▼ ▼ ▼ ▼ ▼ ▼
Claude Code Cursor Codex Gemini OpenCode Your tool
- Architecture Guide — full stack map, 10 IS taxonomy, agent architecture, MCP architecture, infrastructure topology, deployment runbook
- MCP Setup Guide — register the SIS MCP server with every coding agent (Claude Code, Cursor, Windsurf, Codex, Gemini CLI, Antigravity, Hermes) + Railway shared server
- Hermes + Claude Code + OpenClaw Guide — dual-stack setup: Hermes on VPS + Claude Code local + OpenClaw on Railway + phone integration
- Infrastructure Deployment Guide — full 5-surface deployment: Vercel + Railway + VPS + local + phone
git clone https://github.com/frankxai/Starlight-Intelligence-System.git
cd Starlight-Intelligence-System
npm install
npm run build # tsc to dist/
npm test # 82+ orchestrator tests
npm run lint # tsc --noEmitsrc/ is under 3,000 lines of TypeScript with zero runtime dependencies outside better-sqlite3. Substrate docs (SIP.md, SIS.md, etc.) are markdown-only — no build dependency.
- Code (operational layer + substrate-tier commands): MIT — see
LICENSE. - Substrate spec docs (
SIP.md,SIS.md,ALLIANCE.md,STACK.md,VOICES.md,VERTICALS.md,MEMORY.md,REGISTRY.md,SKILL.md): MIT, sameLICENSE. The "Built on SIP" attestation request is a social-layer convention captured inNOTICE, not a license restriction. - Arcanea canon (if your fork composes with it): CC-BY-NC 4.0, © Arcanea BV. Lives in
frankxai/arcanea-ecosystem; canon is compose-only — not redistributed under this repo's MIT. - Trademarks: ARCANEA, FRANKX, STARLIGHT INTELLIGENCE — registered or in registration, reserved rights.
"Built on SIP" is an attestation phrase, not a trademark. Use of the phrase requires actual SIP composition per /sip-attest rules. Full attribution + canon + trademark summary in NOTICE (Apache-style — LICENSE for legal rights, NOTICE for propagation conventions).
- Arcanea — Creator platform; canon-defining vertical built on SIS
- FrankX — Architect brand; SIP thought leadership
- Public Vaults — Browse and fork vaults
- Protocol page — Canonical SIP spec
- GitHub — Source, issues, discussions
Built on SIP · Starlight Intelligence Protocol · v1.1.1 · v8.3.0 · MIT
The product is the repeatable system that delivers a sovereign intelligence estate (personal Mind + 144-agent-class Mesh + Steward) to any principal/alliance/community on SIP. Trinity AI is instance #1.
Positioning: Category owned by Starlight. Program: Olympus. Asset: appreciating, owned intelligence estate — not subscription.
Architecture:
- SIS (substrate: protocol, agent model, catalog, design standard, kernels) — consumed, never forked.
- starlight-estate-os (template repo: scaffold, playbook, generator, module scaffolds, naming profiles).
- estate- (per engagement: tuned contracts, data, fleet, brand voice, modules, CEO assets, private strategy).
The 80/20 reuse boundary (makes margin): Reusable in template/SIS (SIP layers, 4-layer agent model, naming profiles plain/pantheon/luminor/chess, architecture-options catalog + matrix, registry generator, module scaffolds, engagement playbook, offer/PRD/handover templates). Per-client: their data/vaults/prompts, tuned roles, chosen persona+names, selected topology+kernels, fleet config+integrations, tuned module workflows, SOW/pricing/brand voice, filled offer+handover.
Rule: If it isn't theirs specifically, promote to template/SIS so next estate gets free.
Repeatable engagement (one process): Discovery (Mandate) → Blueprint (Mandate) → Pilot (Mind+1st Mesh) → Scale (Full Mesh) → Run (Steward retainer).
Roles in factory: Command center (Cowork) thinks/sequences; Architect/Frank (client rel, blueprint, SOW); Claude Code + claws (build from template+brief); Operator/Steward (provisioning, day-2, evolution). Leverage: Frank sells next while current executes itself.
Tooling: Claude Code (build executor), Cowork (command center + principal Mind), Hermes (local-mesh envelope), SIP (licensed substrate), open standards (AGENTS.md, Skills, MCP), Local LLM (Ollama/MLX on fleet; frontier on command), Design standard (liquid glass, never all-caps, validated).
Economics: Estate #1 amortizes substrate. #2 reuses ~80% → higher margin, faster. Retainer = compounding annuity. Moat = library of reusable estates/modules. Track reuse ruthlessly.
How SIS absorbs: Promote client-agnostic to SIS canon (docs/architecture/, docs/delivery/). Trinity consumes from SIS (licensing clean).
Estate lifecycle: Prospect → Mandate sprint (Phase I, paid) → fork estate-os → configure 4 layers + fork modules → Claude Code builds (Mind → 1st Mesh → full) → handover → Steward retainer → reusable learnings promoted back.
SIP as Web4: Open protocol for self-sovereign intelligence fabrics (like Ethereum for value transfer). Layered: protocol open (MIT, attribution compounds); canon/IP protected (CC-BY-NC, encoded-self non-licensable per SIP §5.7); commercial delivery (factories, retainers, verticals) captures value. Not dumb about open core: adoption flywheel + moat in your unique synthesis (Genius excavation, operated vertical patterns, production harnesses, taste).
See full model, workflow, Trinity split, IP posture, benefits, tech upgrades in the linked strategy + delivery docs above. Board (2026-06-16): PROCEED-WITH-REVISE (REVISE items closed this night — extraction, profile, commands, upgrades track, hero plan, time split).
Trinity management: Alliance layer (governance per ALLIANCE.md, abundant Frank architectural) + separate commercial SOW (scoped build for advanced army using substrate + Frank depth). Explicit time split enforced.
This is how we deliver world-class sovereign agent armies at scale with highest standards.
All foundations verified this night: Harness passed (144 agents, 83 skills, v8.3.0), tests green, SIP/SIS consistent (sovereignty, attestation, 10-IS, encoded-self, no drift), new artifacts integrated with "Built on SIP", excellence on prod branches (main/ship/* clean post-merge/PR).