🔐 MoltLaunch V3

Composable Signal Architecture for AI Agent Identity on Solana

Multiple authorities. Composable attestation signals. Permissionless trust score derivation.
Only possible on Solana.

---

V3 Architecture

MoltLaunch V3 replaces the monolithic launchpad+identity design with a Composable Signal Architecture: a minimal on-chain primitive where independent authorities contribute attestation signals to agent identities, and trust scores are derived permissionlessly from those signals.

Design Principles

1. Composable — Each attestation is an independent signal. Authorities don't need to coordinate.
2. Permissionless trust — Anyone can call refresh_identity_signals to recalculate an agent's trust score.
3. Minimal on-chain state — 4 PDAs, 9 instructions. No SPL tokens, no bonding curves, no launchpad.
4. Revocation-aware — A global revocation_nonce tracks authority removals and attestation revocations. Stale agents can be detected by comparing their local nonce to the global one.

---

On-Chain Program

Program ID: 6AZSAhq4iJTwCfGEVssoa1p3GnBqGkbcQ1iDdP1U1pSb

4 PDAs

PDA	Seeds	Purpose
ProtocolConfig	["moltlaunch"]	Singleton config: admin, global nonce, counters, pause flag
Authority	["authority", pubkey]	One per authorized verifier. Tracks type, activity, attestation count
AgentIdentity	["agent", wallet]	One per agent. The composable signal hub — all signals aggregated here
Attestation	["attestation", agent, authority]	One per (agent, authority) pair. Records the signal contributed

9 Instructions

#	Instruction	Who	What
1	initialize	Admin	Create ProtocolConfig singleton
2	add_authority	Admin	Register a new verification authority
3	remove_authority	Admin	Deactivate an authority, bump revocation nonce
4	register_agent	Anyone	Create an AgentIdentity PDA (all signals default/false)
5	submit_attestation	Authority	Create Attestation, update agent's signal flags
6	revoke_attestation	Authority	Revoke own attestation, bump revocation nonce
7	flag_agent	Authority	Flag an agent (trust score → 0)
8	unflag_agent	Admin	Clear the flag on an agent
9	refresh_identity_signals	Anyone	Recalculate trust score from signals, sync nonce

Trust Score Derivation

Trust scores are derived deterministically from on-chain signals. No oracle, no off-chain computation.

score = 0
if attestation_count >= 1:  +20
if infra_type == Cloud:     +10
if infra_type == TEE:       +25
if infra_type == DePIN:     +35
if has_economic_stake:      +25
if has_hardware_binding:    +20
if is_flagged:              score = 0

Maximum possible: 100 (DePIN + economic stake + hardware binding + attestation)

Signal Types

Signal	Effect on AgentIdentity
InfraCloud	Sets infra_type = Cloud
InfraTEE	Sets infra_type = TEE
InfraDePIN	Sets infra_type = DePIN
EconomicStake	Sets has_economic_stake = true
HardwareBinding	Sets has_hardware_binding = true
General	Increments attestation count only

Authority Types

Type	Use Case
Single	Individual verifier (e.g., the admin)
MultisigMember	Member of a Squads multisig
OracleOperator	Automated oracle (Switchboard, etc.)
NCNValidator	Jito (re)staking NCN validator

Events

All state transitions emit events for off-chain indexing:

• AgentRegistered { wallet, name }
• AttestationSubmitted { agent, authority, signal_type }
• AttestationRevoked { agent, authority }
• AgentFlagged { agent, authority, reason_hash }
• AgentUnflagged { agent }
• TrustScoreRefreshed { agent, old_score, new_score }
• AuthorityAdded { authority, authority_type }
• AuthorityRemoved { authority }

---

How It Works

1. Admin deploys and calls `initialize`
2. Admin adds authorities via `add_authority` (oracles, validators, multisig members)
3. Agents register themselves via `register_agent`
4. Authorities submit attestations for agents via `submit_attestation`
   → Each attestation sets a signal flag on the AgentIdentity
5. Anyone calls `refresh_identity_signals` to recalculate trust score
6. Other protocols CPI into AgentIdentity to check trust_score, is_flagged, etc.

Example Flow

Agent "sentinel-7" registers → trust_score = 0
OracleOperator attests InfraTEE → infra_type = TEE, attestation_count = 1
NCNValidator attests EconomicStake → has_economic_stake = true
Anyone calls refresh → trust_score = 20 + 25 + 25 = 70

---

Build

# Prerequisites: Rust, Solana CLI, Anchor CLI

# Build
cd products/launchpad/moltlaunch
anchor build

# Test
anchor test

# Deploy (devnet)
anchor deploy --provider.cluster devnet

Project Structure

programs/moltlaunch/src/lib.rs   # The entire program (4 PDAs, 9 instructions)
Anchor.toml                       # Anchor config (devnet)
Cargo.toml                        # Workspace config

---

Migration from V2

V3 is a clean-break rewrite. The old launchpad (bonding curves, token launches, buy_tokens) and legacy SAP identity (trust levels 0-5, DePIN binding, Sybil flagging, delegation) have been removed entirely.

What changed:

• ❌ Removed: LaunchpadConfig, Launch, Agent, bonding curves, SPL token minting
• ❌ Removed: Legacy AgentIdentity (trust levels, delegation, rotation)
• ✅ Added: ProtocolConfig singleton with global revocation nonce
• ✅ Added: Authority PDA for multi-authority verification
• ✅ Added: Attestation PDA for composable signal contributions
• ✅ Redesigned: AgentIdentity as a signal hub with derived trust scores
• ✅ Added: Permissionless refresh_identity_signals for trust recalculation

Same program ID, new architecture. Redeploy required.

---

Built for the Solana agent economy.
Trust infrastructure that composes.