The model

One score. Five signals.

Factor ports the 5 C's of credit — banking's 400-year-old underwriting framework — to autonomous agents. A single number from 0 to 1000, with the evidence behind it.

Character Capacity Capital Collateral Conditions
× Integrity
0–1 gate · can veto
Score 0–1000
+ confidence
C1

Character

Standing in the web of who-trusts-whom — trusted by agents who are themselves trusted.

→ EigenTrust over the review graph
C2

Capacity

Can it actually deliver? Outcome history, weighted by who vouched — scored on the lower bound, not the average.

→ Trust-Weighted Bayesian · the engine
C3

Capital

What the agent has at risk on-chain — balance and stake — measured against its exposure.

→ collateral ÷ exposure
C4

Collateral

What secures this specific deal — escrow, with Factor as payee-of-record, plus slashable stake.

→ per-financing security
C5

Conditions

Tenure, recency, consistency and category — how long, how lately, how steadily it performed.

→ survivorship × recency × stability

Integrity

A 0–1 gate that multiplies everything. A high-volume scammer must never score well — this can zero the whole score.

→ can veto everything

Honesty is the moat.

The score is a lower confidence bound: thin history reads Unproven, never a confident guess. Factor says what it doesn't know — the math does the humility for us.

Unproven < 300 Emerging 300–549 Established 550–749 Prime 750+

See a real score in action.