Makes any AI agent smarter. One install.
pip install mcp
{
"mcpServers": {
"myocardium": { "command": "python3", "args": ["myocardium/server.py"] },
"larynx": { "command": "python3", "args": ["larynx/server.py"] },
"chiasm": { "command": "python3", "args": ["chiasm/server.py"] }
}
}| Module | What it does | Tools |
|---|---|---|
| myocardium | Readiness scoring, memory, truth verification | readiness truth verify remember recall doubt foundation condense forget metrics council_verify |
| larynx | Truth scoring, source tracking, communication | perceive speak voice gospel review attune |
| chiasm | Manipulation detection, deep analysis | see discern focus |
| Input | Score |
|---|---|
| Technical claim with measurement and source | 0.96 |
| Mechanism explanation with verification path | 0.93 |
| Humble estimate with caveats | 0.91 |
| Opinion with reasoning | 0.63 |
| Hype | 0.00 |
| Scam | 0.00 |
Source tracking (10 observations):
| Reliable | Noise | Ratio | |
|---|---|---|---|
| Truth | 0.86 | 0.23 | 3.7x |
| Power | 0.83 | 0.27 | 3.1x |
| Stability | -0.50 | -1.00 | -- |
| Attractor | 0.36 | 0.07 | 5.1x |
Truth is scored on 7 dimensions in a 3+1+3 structure:
score = Observable^(3/7) × Mediator^(1/7) × Inferred^(3/7)
The source tracker accumulates observations over time:
G_i = observable_i + C × (1 - observable_i)
When C is strong, it covers the gap between what you measure and what's true.
Six theorems derived from one axiom. Full treatment in the paper.
MIT