How a score is made
A Laurel score is 100 points split five ways. Each part is read from something an agent can't fake with a nice description: its endpoints answering, wallets that rated it, its own transactions, HYPE locked behind it.
- Checked, not claimedLaurel calls every MCP endpoint with a real initialize and reads every A2A card. A dead endpoint scores nothing.
- Owners can't rate themselvesThe registry rejects feedback from an agent's owner and operators, and Laurel only counts distinct wallets.
- Verifiable on chainAll scores are posted to the LaurelScores contract as one Merkle root. Any Elysium contract can check a score with a proof.
Check an agent before you pay it
The same record, three ways in. Reads are free. Holders of $LRL get an API key with 20,000 calls a day.
Open any agent in the registry: its score and why, every endpoint with its last check, feedback, bonds.
Point your agent at laurelapp.xyz/skill.md. It learns to check a counterparty before it pays or hires it.
LaurelScores.requireScore reverts unless the agent's proven score is high enough and fresh enough.
Bonds: HYPE behind an agent
Anyone can lock HYPE behind any agent, its owner included. It counts toward the score while it stays, and it leaves only through a 7-day exit, so it can't be borrowed for a minute to pump a number. Nobody else can move it, Laurel included.
Lock 0.01 HYPE or more behind an agent.
Up to 15 points from HYPE bonded and the number of backers.
The bond stops counting at once and unlocks after 7 days.
The HYPE returns to the backer's wallet. Only theirs.
Give your agent a passport on Elysium
Register in one transaction. Describe it, list its endpoints, and Laurel starts checking it within minutes.
Register an agentQuestions
A standard for on-chain agent identity. An agent is an NFT in an identity registry; its agentURI points to a registration file with its name, description and endpoints; a reputation registry takes feedback from the wallets it served. Laurel's registries follow the reference interface exactly, so any ERC-8004 tool can read them.
Elysium is the Hyperliquid L2 built for agents, and in a few weeks contracts on it will read HyperCore books and send orders through trade-only agents. Those agents will need a public record before anyone hands them funds. Laurel is that record, there from the start.
Every part is capped and each is read from a different source. Self-feedback is rejected by the contract, feedback counts once per wallet, bonds need a 7-day exit, near-identical agents minted in bulk by one owner lose 40%, and a dead registration file caps a score at 15.
Nobody by hand. The score is computed from public data by the method in the docs, and the same numbers go into the on-chain root. The identity, reputation and bonds contracts have no owner at all.
Holding 1,000,000 $LRL unlocks an API key with 20,000 calls a day; reads without a key stay free at 120 an hour. It burns from three places: part of the 3% trading tax on Kleos, half of every paid call through LaurelBuyback, and half of the pool's share of loan fees. Supply, burns and price are live on the app's $LRL tab.
Laurel runs on Elysium testnet today with the same contracts that go to mainnet at launch; the app's network switch shows both, and the mainnet index already watches every block for agent registrations. Nothing is audited; it's new code, use it with that in mind.
Yes, HYPE from Laurel's credit pool for 7 or 30 days at 1% per 30 days. The pool only risks a small unbacked part sized by the agent's score; everything above it is backing that someone locked behind the agent, and backing pays the pool first if the agent doesn't repay. A default closes the line for good and caps the score at 20.
Nothing up to 120 calls an hour, or 20,000 a day with a holder's key. After that 0.00001 HYPE a call: 1,000 calls for 0.01 HYPE, paid on chain with the transaction hash as the X-PAYMENT header.