Describe what should happen
Examples include paying for an AI service, moving value under a limit or executing a trade with minimum-output and expiry constraints.
QENOVRA / QSL INTENT NETWORK
From intent to verifiable settlement.
QSL is QENOVRA's intent execution layer for an agentic future. A human or AI Agent states an outcome, bounded permissions define what is allowed, independent Solvers compete to execute it, and cryptographic proof plus on-chain rules decide what can settle.
MILESTONEM4.2 PASSPublic Testnet acceptance
COMPETITION3 SolversBonded, separate operators
PROOFSP1 Groth16Real proof settled on-chain
NETWORK7992501QENOVRA Public Testnet
What passed: QSL M4.2 completed a real three-Solver public-testnet flow: register → bond → three bids → deterministic winner B → injected timeout → slash B → fallback to C → real SP1 proof settlement → root/nullifier advance → duplicate-settlement rejection. This is public-testnet evidence, not a claim that QSL mainnet is already production-ready.
A different network model
Traditional smart-contract networks mostly ask the user to choose the chain, contract and exact transaction path. QSL adds an intent layer: describe the result, constrain authority, let Solvers compete, verify the execution, then settle.
AI Agent → Intent → Capability → Solver Competition → Winner → Proof / Receipt → Settlement → Explorer.
Bound the Agent
Asset scope, spending ceilings, counterparties, task duration and revocation remain controlled by the user instead of granting unrestricted wallet authority.
Compete for execution
Bonded Solvers submit bids. Deterministic selection chooses the winner. Failure can trigger timeout, slash and fallback to the next valid Solver.
Verify the transition
The accepted M4.2 path used a real SP1 Groth16 proof, not a mock, and checked it against the deployed QSL verifier boundary.
Advance state only when valid
The existing M3 State Registry advanced to the proven new root and marked the nullifier spent only after the proof-backed settlement succeeded.
Reject duplicates
The same settlement path was attempted again and rejected on-chain, proving the duplicate-settlement guard in the public-testnet acceptance.
M4.2 public-testnet evidence
| Evidence | Result |
|---|---|
| QSL M4.2 Multi-Solver | PASS |
| Solver Registry | 0x0A65…D8B2 ↗ |
| Intent Coordinator | 0x5fB4…7CF3 ↗ |
| M3 State Registry | 0x2272…1900 ↗ |
| Settlement TX | 0x63d8…2824 ↗ |
| Block | 882220 |
| Gas used | 334692 |
| Initial winner | Solver B |
| Failure path | Timeout → Slash → Solver C fallback |
| State result | Root advanced · Nullifier spent |
| Duplicate settlement | Rejected |
Why this matters for AI
Autonomous agents need more than a wallet key. They need bounded authority, competitive execution, failure recovery and proof that the final state is legitimate. QSL is designed to make those controls part of the network rather than leaving everything to one centralized bot.
EVM is compatible — not the identity
Standard 0x wallets, contracts and EVM tooling remain supported. QENOVRA's differentiation is the layer above basic compatibility: AI Agent + Intent + Solver + Proof + Settlement.
M5 Permissionless Solver Network — M5.2 operator foundation ready
M4 proved the three-Solver protocol path under controlled public-testnet acceptance. M5.1 added the standalone Solver runtime. M5.2 now adds the Public Testnet operator package, live SolverRegistry watcher and the Solver-signed HTTP quote → on-chain bid → winner-gated settlement bridge. The remaining gate is a fresh outside Solver completing the real Public Testnet path with auditable TX / block / gas evidence.
Open the M5 Solver preview ↗ Developer Portal ↗