AI BVF
Open protocol for scoring AI investments. Dual-licensed: grammar under CC-BY-4.0, code under MIT, benchmark corpus and certification marks proprietary.
Packages
| Package | Install | Purpose |
|---|---|---|
@aibvf/core | npm i @aibvf/core | JS/TS validator + scoring engine |
aibvf | pip install aibvf | Python validator + scoring engine |
aibvf-mcp | npx aibvf-mcp | MCP server: makes the scorer callable from any Claude agent |
All three ship the same scoring logic and validate against the same JSON Schema.
Quick use
JavaScript
import { score, validate } from '@aibvf/core';
const result = score({
industry: 'manufacturing',
revenue_eur: 2_400_000_000,
function: 'supply',
ai_tier: 'gen2',
readiness: 'traditional',
scores: {
strategic_alignment: 72,
financial_return: 64,
change_enablement: 48,
governance_risk: 35,
},
});
// => { classification: 'Fix', net_low_eur: 75_600_000, net_high_eur: 247_000_000, ... }
Python
from aibvf import score, validate
result = score(
industry='manufacturing',
revenue_eur=2_400_000_000,
function='supply',
ai_tier='gen2',
readiness='traditional',
scores={'strategic_alignment': 72, 'financial_return': 64,
'change_enablement': 48, 'governance_risk': 35},
)
MCP (from Claude Desktop / Cursor / any MCP host)
Add to your MCP config:
{
"mcpServers": {
"aibvf": { "command": "npx", "args": ["-y", "aibvf-mcp"] }
}
}
Then ask Claude: "Score this AI initiative using AI BVF" and Claude will call the tool.
Publishing
# JS + MCP
npm install
npm -w @aibvf/core run build
npm -w @aibvf/core publish --access public
npm -w aibvf-mcp run build
npm -w aibvf-mcp publish --access public
# Python
cd packages/py
python -m build
twine upload dist/*
Spec
bvf-protocol.schema.json is the canonical JSON Schema. The narrative specification lives at bvf-app.vercel.app/protocol.
License
- Code (all packages): MIT
- Spec text (
./spec/): CC-BY-4.0 - AI BVF, AI BVF Certified, logo: trademark, not licensed under either of the above.