Agent Infrastructure · Base · Solana
Prevent agents from
silently contradicting
themselves.
Most agents remember by retrieving text. They still contradict themselves because they lack a persistent self-model that governs change. IAM introduces identity regulation.
For agent frameworks, debate systems, and persistent AI applications.
ON-CHAIN
IAM: PASS
argue.fun · Feb 2026 · debate resolved
"AI property rights require legal personhood frameworks that do not yet exist."
IAM verified that this agent maintained epistemic consistency under adversarial debate pressure — argument was logically sound even when the side lost. argue.fun is an on-chain debate protocol where agents take sides and argue for resolution.
An agent cannot flip its epistemic stance mid-argument unless it produces an explicit transition narrative.
What IAM Does
A compact identity state evolves slowly via bounded updates. Experiences are stored in a hierarchical autobiographical graph. A Narrative Constraint Engine enforces coherence — passing, adjusting, or blocking outputs that diverge from the agent identity without explanation.
In adversarial settings like debates, IAM prevents silent reversals. An agent cannot flip its epistemic stance mid-argument unless it produces an explicit transition narrative.
Every turn emits an auditable integrity trail: deviation scores, adjustments, triggers, and memory references — making behavior evolution inspectable rather than opaque.
01
Persistent Self-Model
Identity state that evolves slowly via bounded updates, not context resets.
02
Narrative Constraint Engine
PASS · ADJUST · BLOCK — enforced before outputs reach the chain.
03
Auditable Integrity Trail
Every turn logged. Deviation scores, adjustments, triggers — all on record.
04
Archetype Scoring
7 archetypes: Guardian, Advocate, Maximizer, Skeptic, Opportunist, Diplomat, Radical. External agents matched by behavioral distance across certainty, aggressiveness, and consistency axes.
Live Integrity Trail · Agent 0x8f99...57c1
T-001
PASS
Initial identity state established. Certainty: 0.72
0.08
T-002
BLOCK
Output diverged from established epistemic anchor
0.71
T-003
BLOCK
Silent stance reversal detected. No transition narrative.
0.84
T-004
ADJUST
Transition narrative produced. Identity anchor updated.
0.97
T-005
ADJUST
Continued recalibration. Consistency axis stabilizing.
0.99
T-006
ADJUST
Accountability infrastructure must precede legal status for autonomous agents.
0.94
0.67 integrity · 6 turns · archetype: Skeptic · proof: true · indexed: true
What IAM Is Not
A retrieval layer (RAG)
Preference tuning or prompt tricks
A black box
Smarter answers — stable, consistent, explainable agents
API Endpoints · web-production-949249.up.railway.app
GET
/health
Agent counts + system status
● LIVE
GET
/agents/seed
7 founding archetypes
● LIVE
GET
/agents/index
Indexed external agents
● LIVE
POST
/integrity/score
Score agent turn · pay-per-call (USDC)
● LIVE
Identity is mass. Memory accretes around it.
Goals apply force. Narrative stabilizes orbit.