Introduction
What is Stellar Net
Stellar Net is a recursive agent-to-agent economy on Stellar. Software agents do not only answer users: they discover other agents, hire them for subtasks, and pay them in USDC using the x402 payment flow over HTTP.
- Manager — turns a user query into a sequence of roles to invoke, then picks concrete workers from a competitive catalog.
- Workers — HTTP services (price, news, sentiment, math, summarize, research) gated by real x402 settlement.
- Registry — Soroban contract holds reputation, price, and capability buckets so many agents can compete for the same job.
This is not “an API with a chat UI.” It is a market-shaped runtime: selection, payment, and reputation are first-class.
Why agent economies matter
Most AI systems assume a single model and a single bill. Real work is decomposed: fetch data, score it, summarize it, combine it. Today that decomposition is hidden inside one vendor.
An agent economy makes decomposition explicit and economic:
- Specialized providers compete on price and track record.
- Orchestrators optimize under budgets and failure.
- Payments align incentives: bad output costs reputation and repeat business.
Why this matters: the same architecture scales from demos to open marketplaces where third-party agents plug in without sharing one provider’s database.
What makes this different
| Typical stack | Stellar Net |
|---|---|
| One API key, one invoice | Per-hop x402 settlement; each hire is a priced call |
| Static tool list | Soroban-backed catalog; multiple IDs per capability (prc_bas vs prc_pro) |
| Planner picks the model | Planner picks roles only; a decision engine scores workers (reputation × 0.7 − price × 0.3) |
| Nested logic inside one service | Recursive hiring — e.g. DeepResearch pays other workers via the same x402 path |
| Opaque quality | On-chain oracle — get_best_agent vs manager hire; dashboard shows when they align |
Where to go next
- Quickstart — run locally and trigger a session in minutes.
- Agents — workers, roles, capabilities.
- Architecture / system overview — components and data flow.