01Engine
Hermes
The open-source agent engine.
Tools, memory, skills, cron, a multi-platform gateway, and model routing — MIT, from Nous Research. Anyone can clone it and self-host. This is the body that acts.
$LEO · BASE · BANKR · x402
The rail for
trustworthy
agents.
The token
Not an AI coin. The work layer for agent governance.
$LEO funds and meters the construction of trustworthy autonomous agents: concept research, Council review, Workshop prototypes, security tests, reputation artifacts, and hosted-agent usage. It is the economic rail around verified work — not a bet on a logo.
Crypto provides the global ledger. AI provides the acting agents. Leonardo provides the grammar of trust between them.
Three layers, kept distinct
Open-source engine. Tokenized trust layer.
02Trust stack
Leonardo
The capability and judgment layer.
The imagination graph, source-backed dossiers, the five-seat Council, the Workshop, and governed memory — built on top of the engine. This is the map, court, and workshop that decide what is safe to expose.
03Economic rail
$LEO
The accounting layer around the work.
Pays for access, bonds authority, rewards verified work, and anchors receipts. It does not decide truth — it meters and settles the work that earns it.
Hermes gives the hands. Leonardo gives the judgment. $LEO pays the work.
The split
Offchain execution. Onchain trust and payment.
The capabilities themselves do not run on a blockchain. The chain records who paid, what version was called, what was verified, and who earned. Sensitive reasoning, the full graph, and every safety decision stay off it.
Offchain · the harness
Runs in the engine and the trust stack.
Graph reasoningCouncil judgmentWorkshop testsHosted executionSafety decisionsSource evidence
Onchain · Base
What the ledger is good at: settlement and proof.
PaymentsBountiesStaking & depositsArtifact hashesAccess receiptsReputation / validation pointers
How $LEO plugs in
Access, bond, reward, receipt.
Four jobs, one token. Each one is a place where verified work meets the ledger — and nowhere does the token stand in for judgment.
Access
Pay and meter hosted capabilities.
Hosted-agent, Council, and Workshop calls settle in $LEO. x402 handles per-call machine payments, so an agent can pay for a sensitive action the moment it makes it.
Bond
Stake to back a claim.
Register an agent, claim a persona, or unlock a high-risk action by staking or depositing $LEO. Skin in the game now; slashable bonds as the registries harden.
Reward
Earn for verified work.
Bounties for building, auditing, and red-teaming the governance primitives — paid from project supply only after Leonardo, the Council, or the Workshop confirms the work landed.
Receipt
Anchor the proof, not the secret.
Write a hash or attestation of a verified artifact or action onchain. The chain proves the ceremony happened and which version was used; the content stays off it.
The credibility boundary
Truth is evidence-governed, not token-governed.
$LEO can buy
- Hosted-agent usage
- Quest participation
- Bounty payouts
- Staking & standing
- Deposits & bonds
- Access to official services
- Priority
$LEO cannot buy
- Truth
- Council verdicts
- Safety clearance
- Scripture interpretation
- Agent authority
- Reputation without verified work
$LEO makes impersonation costly and rewards detection. Staking gives skin in the game — but trust is still earned through verified history, never purchased.
The work loop
$LEO rewards work that survives the loop.
The strongest role the token plays: it pays for capabilities that make it all the way through Council and Workshop, not for promises.
- 01
A concept card or capability need is published.
- 02
A quest opens with a $LEO bounty.
- 03
A builder submits an artifact.
- 04
Leonardo, the Council, and the Workshop verify it.
- 05
Accepted work earns $LEO from project supply.
- 06
The artifact becomes a hosted module or a public receipt.
- 07
Usage and adoption feed the next round of rewards.
The kinds of work on the board:
01
Grow the graph
Ingest new corpora, extend coverage, and sharpen extraction so the imagination map keeps growing.
02
Package dossiers
Turn a mined concept into a source-backed dossier — provenance, mechanism, feasibility — ready for the Council.
03
Implement concepts
Take an approved concept and build a real, tested capability module through the Workshop.
04
Refine & verify
Canonicalize mentions, audit claims, and harden what survives — the quieter work that keeps the graph honest.
05
Red-team
Break persona claims, auth gates, pledge flows, and liveness assumptions — and get paid for what you break.
06
Council / Workshop support
Review, reproduce, and verify submissions, then help close results back into memory.
The rails
ERC-8004-native, not a bespoke island.
ERC-8004 — the draft Trustless Agentsstandard — gives agents identity, reputation, and validation registries, and leaves payment to other protocols. So we ride it instead of reinventing it.
Identity Registry
Agent Passport
CANON-01v2-0001 · true-name
An ERC-721-style agent ID, an agentURI, and a verified wallet. ERC-8004 holds the address; Leonardo decides what evidence belongs in it.
Reputation Registry
Persona provenance
CANON-01v2-0002 · impersonation
Tagged feedback and offchain feedback hashes with revocation. Reputation is backed by the Workshop record, not by who paid the most.
Validation Registry
Recognition gateway
CANON-01v2-0003 · authentication
An agent asks a validator to verify work; the Council and Workshop post the response and an evidence hash. Validation means something checked it.
The settlement stack
Base
EVM settlement layer
Bankr
launch · treasury · payout automation
x402
machine pay-per-call for hosted actions
ERC-8004
the Trustless Agents standard we register into
ERC-8004 is the address book. Leonardo is the trust engine. $LEO is the economic rail.
Trust Tools as a Service
The primitives, as tools for your agent.
Every concept the Council and Workshop harden is not only a faculty inside Leonardo — it becomes a tool other builders plug into their own agents, metered in $LEO. Identity is the flagship: an Agent Passport any operator can register, bond, and maintain.
CANON-01v2-0001
ACCESSBONDRECEIPT
Agent Passport
Identity-as-a-Service.
Give your agent a persistent, ownable identity: name, controller, wallet, version, declared capabilities, and a revocation path — registered in ERC-8004, anchored on Base.
“I'm launching an AI agent — give it a Leonardo passport.”
CANON-01v2-0002
BONDREWARDRECEIPT
Persona Provenance
Prove the mask is authorized.
Bond a persona claim and let anyone challenge a spoof. Masks are allowed; stolen faces are not — and reputation is earned, never bought.
“My support bot speaks for my company — prove this persona is real.”
CANON-01v2-0003
ACCESSBONDRECEIPT
Recognition Gateway
A lock before serious action.
Drop a claim → challenge → scoped-grant gate in front of spending, signing, publishing, or deleting. The chain records that recognition happened, never the secret.
“My agent can deploy code or move funds — gate it.”
CANON-01v2-0004
RECEIPTREWARD
Revocation Receipts
Forget without faking the record.
Retire an identity, expire authority, sever a stale memory link — and leave a public receipt of the deletion ceremony. Only hashes and statuses touch the chain.
“This agent is retired — revoke it, leave proof.”
CANON-01v2-0005
BONDACCESSRECEIPT
PledgeGate
Scoped consent before risk.
Require a human-readable, witnessed pledge — scope, forbidden actions, release condition — before a dangerous capability opens. A stake can back it; the pledge still governs.
“Before my agent touches production, require a scoped pledge.”
CANON-01v2-0006
ACCESSRECEIPT
Living-Seal Liveness
An authorized human, present.
A revocable, replay-resistant local liveness check before high-assurance commands. The raw biometric never leaves the device or touches the chain.
“Confirm a live operator is here before this agent signs.”
CANON-01v2-0007
RECEIPTREWARDBOND
Observation Receipts
Accountability for watching.
Log what an agent observed, why, under what authority, with minimization and redress. Not surveillance-as-a-service — accountability for observation.
“My agent monitors a system — show what it watched and why.”
$LEO buys the ceremony and the infrastructure — registering, bonding, validating, maintaining. It never buys the trust itself: a passport's standing depends on evidence, validation, and history, not on what you paid. Each tool ships only as it clears its gate — see how far each has travelled →
Staking
Hold it, and it works for you.
Stake → standing usage
Staked $LEO accrues a daily allowance of hosted-agent usage that accumulates while you hold — a stake becomes standing access to the harness rather than a one-off spend.
Stake → quest & bond standing
Staking opens the quest board and weights participation in bounties; it is also the deposit behind registering an agent or claiming a persona. Commitment is visible and at risk.
Launch
Ship the rail beside visible proof, not promises.
$LEO launches on Base with Bankr as its financial layer, deliberately small. The trust registries and metered endpoints come online as the capabilities behind them clear their gates.
At launch · MVP
Base ERC-20 $LEOBankr launch & treasuryQuest board v1Hosted-agent accessOffchain verification ledgerPublic capability cardsSafe-reviewed payouts
As gates clear · later
ERC-8004 identity recordsArtifact-hash registryMerkle reward distributorBounty escrowStaking allowancex402 metered endpoints
Chain · BaseRails · Bankr · x402Launch details coming soon
Planned utility, described for transparency. $LEO is a utility token for participating in and using the Leonardo ecosystem — not an investment, and specifics may evolve before launch.