Solana agents deserve a proof-native control surface.
SWARM turns a claw-machine style agent demo into a polished product showcase: Phantom identity, autonomous execution, memory, and judge-readable proof flow in one coherent cyberpunk narrative.
Phantom identity primes the run
The user starts with Phantom as the explicit Solana adapter, then signs a human-readable session message before the demo narrative advances.
Explicit Phantom adapter
Agent skill is selected with intent
The claw-machine agent chooses a target skill, explains why the action matters, and packages the instruction as a replayable mission card.
Skill manifest mock hash
Autonomous execution produces a receipt
The run simulates a deterministic agent loop that turns wallet intent into a task result, a receipt trail, and a chain-ready outcome.
Receipt SWARM-CLAW-042
Memory and reflection close the loop
The product records what happened, why it happened, and what the agent should remember before the next Solana-linked action.
Memory capsule committed
Judges see a clean proof story
The interface compresses the full wallet → skill → execution → memory → Solana proof path into a single judge-readable sequence.
Explorer-ready proof rail
How It Works
One wallet-gated story replaces scattered command panels.
The application presents SWARM as a product a judge can understand in sequence. The same mock data source powers the landing preview, dashboard narrative, and submission evidence so the experience stays consistent across routes.
Phantom-first wallet call-to-action on every route
Linear story dashboard built from one coherent mock data source
Judge-facing SWARM rubric mapping with concrete product evidence
Deployment-ready Manus static asset paths and route structure