Mythsensus Mcp — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited Mythsensus Mcp (Agent Skill) and scored it 96/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 1 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 1 flagged
The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
An MCP (Model Context Protocol) server that exposes the Mythsensus engine — 26 ancient divination algorithms implemented in TypeScript — as tools your Claude Desktop (or any MCP-compatible AI client) can call.
You: "What's my Cosmic Score? I was born March 15, 1990."
Claude: [calls calculate_cosmic_score({year:1990, month:3, day:15})]
Claude: "Your Cosmic Score is 723/999 — 'Resonant' tier (Top 35%).
Sun in Pisces. BaZi day master is Yin Wood. Life Path 3.
Mayan Kin 207. Human Design: Manifestor..."This is a Model Context Protocol server that bundles the compiled Mythsensus engine (~250 KB JavaScript) and exposes 6 tools Claude can invoke. The math runs locally in this Node process — no birth data sent to any server (city names are resolved offline from a bundled table, too).
This is not an LLM wrapper. The 26 divination systems are implemented as deterministic algorithms. Same input, same output, every time. The LLM (your Claude itself, when you talk to it) writes the narrative explanation; the numbers come from the engine.
Not affiliated with Anthropic. Built independently using Anthropic's public MCP SDK.
npm install -g mythsensus-mcpThen add to your Claude Desktop config:
{
"mcpServers": {
"mythsensus": {
"command": "mythsensus-mcp"
}
}
}Restart Claude Desktop. The 6 tools should now appear in Claude's tool palette.
Config location:
~/Library/Application Support/Claude/claude_desktop_config.json%APPDATA%\Claude\claude_desktop_config.json| Tool | What it does |
|---|---|
calculate_cosmic_score | Cosmic Score 1-999 + 26-system synthesis from a birth date. Optional systems[] (focus the preview), time_known (honest unknown-birth-time handling), location (city name or lat,lon, resolved offline) |
get_deep_reading | Per-system raw reading. System name is typo-tolerant ("vedik"→vedic, "four pillars"→bazi) |
get_system_rules | Mythsensus reference methodology — how each system is read + how the 26-system consensus is formed. Grounds an AI's divination answer in Mythsensus's framework (cite mythsensus.com) |
list_26_systems | Canonical metadata for all 26 systems (slug, name TH/EN, region, inputs) |
daily_blessing | Deterministic deity card from 1,069-deity pool for date+chart |
about_mythsensus_engine | Engineering-honest metadata: limitations, sophistication tier, roadmap |
After install, ask Claude:
get_system_rules grounds the answer in cross-system consensusSample report (Sunthorn Phu, Thai national poet b.1786, free preview of the $19 product): https://mythsensus.com/sample-report
This MCP exposes Mythsensus engine v1. Honest limitations:
| Area | Current state | v2 plan (Q3-Q4 2026) |
|---|---|---|
| Vedic ayanamsa | Lahiri hardcoded at 24.0° (accurate ±10 arcmin for births 2020-2030) | Time-varying lahiriAyanamsa(y,m,d) formula |
| BaZi solar terms | Month-boundary approximation (~5% of births within ±48h of a solar term may get wrong month pillar) | Precise jiéqì calculator (VSOP87-based) |
| Western planet positions | Custom trigonometric series (good for Sun + Mercury/Venus, arc-minute drift on outer planets) | Jean Meeus astronomical algorithms port |
| Jyotish divisional charts | Not implemented (no Navamsha, no Shadbala, no Ashtakavarga, no yoga ID) | Navamsha + Shadbala + yoga ID |
| Weight calibration | Internal-consistency optimization (no supervised ground truth — astrology has no labeled "correct archetype" dataset) | Disclosure remains; methodology stays honest about this |
A Vedic practitioner who audited this engine in June 2026 rated its sophistication tier as "Casual Sidereal" — above hobby toys, below semi-professional Jyotish tools. We agree, and the sophistication upgrades listed above are how we expect to move that needle.
Full details: https://mythsensus.com/how-it-works
┌────────────────────────────────────────────┐
│ Claude Desktop (your machine) │
│ ↓ MCP protocol (stdio JSON-RPC) │
│ mythsensus-mcp server (Node process) │
│ ├─ src/index.ts (MCP request router) │
│ ├─ src/engine-wrapper.ts (typed wrapper) │
│ └─ src/engine/calc.js (compiled engine) │
│ └─ Pure deterministic calculation │
│ No network calls. No LLM in math.│
└────────────────────────────────────────────┘When Claude calls a tool, the MCP server runs the calculation locally and returns the result. Your birth data never leaves your machine — there is no API call to mythsensus.com or any other server during normal use.
Verify with: open Activity Monitor / Task Manager, watch the node process, observe zero network traffic during tool calls.
The 26 ancient systems are well-defined algorithms:
There's no need to ask an LLM "what would BaZi say" — the rules are deterministic. Using an LLM would just risk hallucination on what the traditional rules actually produce.
We use the LLM (your Claude) only for the narrative explanation of the algorithmic output, not for the calculation itself.
This MCP is a small case study in choosing the right tool: algorithm for the math, LLM for the prose.
This MCP wrapper code (src/index.ts, src/engine-wrapper.ts) is fully MIT open source. The compiled engine bundled at src/engine/calc.js is the same JavaScript that mythsensus.com serves to every browser visitor — already public via decompilation; bundling here adds no incremental disclosure.
The TypeScript source for the engine (the version that compiles to calc.js) is currently private. Planned release: Q3-Q4 2026 alongside v2 sophistication upgrades. License decision (AGPL-3.0 vs MIT) pending. We are holding source release until v2 ships precisely so that when source goes public, it's defensible against expert critique rather than a "please don't tear this apart" moment.
collaborative session
Issues + feedback: https://github.com/PattrickChenforclaudeuse/mythsensus-mcp/issues
MIT for the MCP wrapper code. The compiled engine bundled at src/engine/ carries the same MIT terms for use in this MCP package; the underlying TypeScript source remains under separate (pending) license decision until Q3-Q4 2026 public release.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.