MVP Demo

End-to-End Flow

The complete ResearchAgent scenario — three distinct phases showing how a consumer agent autonomously purchases premium data from multiple sources on behalf of a user who set a $10/day budget once and never touched the app again.

The Demo Scenario

The MVP demonstrates a complete AP2 → REAPP → x402 → Stellar flow using a research data scenario. The scenario was chosen deliberately: it exercises every component of the stack simultaneously — mandate issuance, scope validation, Soroban enforcement, x402 settlement, fee sponsorship, multi-merchant composability, and audit trail — in a single coherent user story grant reviewers can follow.

Three distinct phases. The user touches the system exactly once. Everything after is fully autonomous.

Phase 1 — Setup

The user opens the REAPP web wallet, signs an IntentMandate with their Freighter desktop wallet, and registers it on Soroban. The agent keypair is added as a scoped policy signer. This happens once. The user's private key is never involved in any subsequent operation.

Phase 1 — Setup (One-Time User Action)
The user's private key is involved exactly once — at mandate registration. After setup completes, the agent operates fully autonomously within the mandate bounds.

Phase 2 & 3 — Autonomous Operation + Settlement

The agent receives a task, identifies data sources, and executes the full validation-to-settlement cycle without any user involvement. The REAPP validator runs five checks before any signing happens. Stellar settles in approximately three seconds. The Soroban period counter updates atomically with each payment.

Phase 2 & 3 — Autonomous Operation + Settlement
From task receipt to report delivery. The user's key is never touched. The agent signs only scoped auth entries. LaunchTube sponsors the XLM fee. The full audit trail is written on-chain.

Multi-Source Budget Tracking

The same IntentMandate covers multiple fulfillment agents in a single session. After purchasing from three separate merchants — DataFeed.xyz, ResearchAPI.io, and NewsFilter.ai — the agent has consumed 4.00 of its 10.00 USDC daily budget across three Stellar transactions. The Soroban contract tracks the aggregate spend atomically. The agent knows exactly how much budget remains.

Multi-Source Budget Tracking
One mandate, three merchants, three Stellar transactions, one shared period counter. The Soroban contract enforces the aggregate daily budget — not each transaction in isolation.

What the MVP Actually Proves

✅No key custody
Agent uses scoped auth entry signing. User key never leaves Freighter hardware wallet during any payment operation.
✅On-chain spending limits
Soroban contract enforces daily budget atomically. Overspend attempt fails at the contract — the agent cannot override this.
✅AP2 mandate validation
Cryptographically signed mandates correctly scope agent authority. Merchant list, amount limits, and expiry enforced before signing.
✅x402 compatibility
Full HTTP 402 flow with x402stellar packages. verify + settle cycle works end-to-end on Stellar testnet.
✅2–5 second settlement
USDC transfers on Stellar testnet with LaunchTube fee sponsorship. Fixed $0.00001 fee regardless of payment amount.
✅Mandate audit trail
Every payment atomically links mandate_hash to tx_hash in Soroban registry. Full chain reconstructable on-chain.
✅Multi-merchant composability
Same mandate, three different merchants, three separate Stellar transactions, shared period counter.
✅Evaluator pattern
Deferred task payments with Soroban escrow and three-tier evaluator attestation. Interface specified and reference implementation shipped.

Build Reality — What Already Exists vs Net-New

A lot already exists. x402stellar's four npm packages are live. Soroban custom account patterns are documented. The AP2 spec and Python reference library are published. The team is wrapping and composing — not building from scratch. The real complexity is the validator and the Soroban contract, not the SDK surface area.

The honest timeline assessment: 5 months, 2 engineers, ~800 hours total is tight but doable if AP2 stays stable. Scope creep or AP2 instability are the most likely timeline threats.

What Exists vs What REAPP Builds Net-New
The SDK packages are largely thin wrappers once the validator and Soroban contract work. The adapter-first validator architecture is the highest-complexity deliverable.
← PreviousAgent Patterns
Next →Deployment