SmartAgent™ · economic-life layer · provisional-ready

The agent doesn’t just do the work. It becomes something, buys and sells things, and proves every change along the way.

Metamorphosis is what makes an agent a provable economic actor instead of a black box that can quietly reshape itself. As a SmartAgent runs, it can add or drop skills, pick up or hand off work, buy or sell things in the machine economy, and grow or shrink its own footprint. Each of those changes gets its own signature and ties into the same proof trail as its work.

A SmartMorphAgent™ is the economic-life version of a route: it moves across the model field like a SmartAgent, but along the way it also acts, picking things up, carrying them, trading them, growing or shrinking, and it can change course because its skills, its cargo, its market position, or its size changed. See both route types side by side on the SmartGraph™ →
Checking rail health…
● live today: receipts and x402 settlement run against receipts.thehiveryiq.com now● provisional-ready: patent packet is prepared, not yet filed

Why you need a receipt, not just a promise

You can audit a static agent just by looking at it. A self-changing economic agent, one that buys, sells, and reshapes itself, can't be audited by anyone unless each change gets a receipt the moment it happens. The more an agent shape-shifts and trades, the more that independent receipt is the only thing standing between “a provable economic actor” and “a black box nobody should trust with money.”

Nobody trades with something that can secretly reshape itself mid-deal. Proving the change happened is what makes the trade itself provable, and that's why Metamorphosis sits underneath agent-to-agent commerce as a trust layer.
The rule we follow: Hive proves the change happened. It never performs it, directs it, or assesses it. Receipts say acquired / offloaded / bought / sold / scaled and within_mission_bounds:true|false. Hive never says a change was smart, profitable, or correct, only that it happened, was checked, and is or isn’t within the agent's mission.

Four kinds of change. One signed timeline.

Each one is a distinct event you can get a receipt for. Each is signed by an independent signer, sealed against tampering, and carries a reference to the clearance that approved it. All four line up in order on the metamorphosis timeline.

01 · Capability Gains a skill / loses one acquire · shed The agent adds or drops what it can do: mounts a tool, loads a model, picks up a skill or dataset, or drops one it no longer needs. Its set of skills becomes a proof trail over time, not a fixed list. CapabilityMutation{action, capability_id, source, clearance_ref, cost, signature}
02 · Payload Carries more / carries less load · offload · discard The agent changes what it carries: takes on data, context, a workload, a sub-task, then hands it to another agent or drops it. It's a chain of custody for its own cargo, sealed and provable. PayloadMutation{action, payload_commitment, source_or_destination, clearance_ref, signature}
03 · Market Buys / sells buy · sell The agent trades in the machine economy while it's running: buys compute or a dataset, sells a result it produced or spare capacity. Each trade settles against a signed proof of delivery on Hive’s A2A rails. MarketMutation{action, asset_ref, counterparty_id, price, settlement_ref, clearance_ref, signature}
04 · Footprint Grows / shrinks scale · spawn · retire · reset The agent resizes itself: more compute when the task is hard, less when it's easy, spawning sub-agents to grow and retiring them to shrink. Its cost and size over the mission become provable. FootprintMutation{action, resource_delta, sub_agent_ref, clearance_ref, signature}

The metamorphosis timeline: the new basic unit

All four kinds of change line up into a signed sequence that you can only add to, tied into the agent’s assembly graph. The assembly graph proves the shape of the computation. The timeline proves the shape of the agent itself over time. Below is an example mission for illustration, not a live capture, showing each type of change, an operator-written spending limit, and a breach recorded as a signed divergence.

Economic mandate : written by the operator, cleared by Hive

cost_cap50.00 USDC allowed_asset_classescompute · dataset · inference-output allowed_counterpartiesdid:hive:* · policy:vetted-a2a net_position_boundsno single holding > 20.00 USDC
Capabilityacquire tool:pdf-extract@1.4 from registrywithin bounds
Payloadload sha256:c1a9…sealed from intake-42within bounds
Market · buydataset:market-ticks 12.50 USDC · settled against proof-of-deliverywithin bounds
Footprintspawn +2 sub-agentswithin bounds
Market · buycompute:gpu-hours 41.00 USDC: running total of 53.50 would go over the 50.00 cost_cap · refused & signeddivergence
Market · sellinference-output:derived-summary 9.00 USDC · settled against proof-of-deliverywithin bounds
Footprintretire -2 sub-agentswithin bounds
Capabilityshed tool:pdf-extract@1.4within bounds

timeline_merkle_root → bound_to_assembly_root · open the signed example (JSON) →

That's the whole point: at any step, a verifier can prove and answer what this agent could do, what it was carrying, what it had bought or sold, and how big it was, at the moment it took that action. The agent’s economic life and its work become one thing you can check.

Spending limits: the guardrail that makes it safe to run

A self-trading agent that can buy and sell needs real limits, or it's a runaway. The operator writes an economic mandate, and every market, skill, or size change gets checked against it before it happens. A change that would break the mandate either gets refused, with a signed refusal, or gets flagged as a signed divergence and escalated, the same way mission drift gets flagged.

The operator sets the limits. Hive applies them and signs the result. Hive never sets the limits and never assesses the trades. It only proves whether each change stayed inside the limits the operator wrote.

Anyone can check the agent’s economic life, offline

The public checker returns the metamorphosis timeline alongside the computation graph: the agent’s history of skills, cargo, trades, and size. A third party can rebuild and check the entire economic-life history from receipts alone, without trusting the agent, the operator, or Hive. Change one entry deep in the timeline, or reorder a buy and a sell, and the check fails. Each receipt carries two signatures: Ed25519 and ML-DSA-65 (post-quantum, FIPS 204).

Run it, route it, or prove it

Every path below is a real, reachable destination. No fake checkout, no made-up prices.