What's built, what's funded, what's next.

The three spec endpoints, running against Stellar testnet, callable right now.

Live endpoint facilitator deployment down

Includes areFeesSponsored: true. Byte-identical means a stock client needs no walras-specific branch.

FACTS F-041 ↗

Accounts via Friendbot, testnet USDC acquired on the DEX at run time, zero stored secrets. The client is stock — no fork, no patch, no shim.

CI run ↗settled tx ↗

Measured across CI runs, not projected. Replace with live aggregates once the observatory exists.

EVIDENCE §S7-2 ↗

The gate fails the build if a copyleft dependency enters the tree.

GitHub ↗

FACTS: every protocol claim dated and source-pinned. EVIDENCE: a transcript or a hash for every capability claim.

Repo ↗

Channel-account pool for bursty agent traffic; idempotency and settle-retry with hash recovery; boot-time fail-closed invariants; full negative-path matrix (malformed XDR, wrong network / asset / amount / recipient, expired, replayed, sponsorship-safety violations) each returning a machine-readable, non-null reason.

HTTP and MCP resources keyed on the spec tuple; schema soft-drop; routeTemplate percent-decoded before traversal checks; listings bound to the settled payment's recipient; EXTENSION-RESPONSES outcomes.

The spec's six filters, stable cursor pagination.

One-line, framework-agnostic wrapper for discovery metadata (per-parameter descriptions), plus an LLM onboarding prompt that interviews a seller and generates their complete discovery manifest.

A real Policywright tool priced in testnet USDC through walras, auto-cataloged by its first payment, keyed on (resource.url, input.toolName).

Hybrid lexical + semantic retrieval behind a stable Retriever interface; cursor pagination and partialResults per spec.

Golden query set built from the real catalog with graded judgments; nDCG@10, MRR, recall@k published and enforced as a CI regression gate that demonstrably fails on a deliberately regressing commit.

scheme_upto_stellar.md authored; minimal Soroban settlement contract (no admin, no upgrade path, no custody, no persistent balance) enforcing recipient binding, single settlement, ceiling and ledger time-bound via require_auth_for_args; facilitator-side validation; composition with smart-account spending policies documented; upstream PR opened and coordinated through the x402 TSC.

Every attack surface — payment forgery, catalog poisoning, listing spoofing, fee drain, replay — mapped to its control and its test.

Alerts demonstrated firing against the live testnet deployment.

search_resources and paid_call, deterministic schemas, one error vocabulary. Signing stays client-side; the server holds no keys.

Audit Bank findings resolved; fail-closed pubnet configuration.

exact and upto live on stellar:pubnet; conformance re-run on both networks; settled hash published per network per scheme.

Depends on: Security review remediation

Role-based — seller / buyer & agent / operator — with live testnet examples.

A paid API discovered and paid by an agent; an MCP agent paying with no pre-baked integration. Both runnable from a clean clone.

Helpers plus a guide, with one worked port of an existing Base/Solana x402 seller adding stellar:pubnet at config level.

Runbook; live monitoring and public status page (aggregate metrics only, no PII); degraded-mode behavior documented; spec-change tracking with a named maintainer and a stated conformance-update cadence through and beyond the grant.

Per SCF requirement: a first-time developer measured from docs to a paid, discoverable endpoint in under an hour.

Reverse proxy in front of any existing API; payment-gate a route with zero code changes. A self-serve version of what Cloudflare has behind an early-access waitlist.

Revenue, per-resource analytics, request/settle funnel, error taxonomy. Reads the same public endpoints agents read.

Every settlement emits a verifiable receipt; webhook on settle and on dispute-window close. Receipts become the substrate for Layer 3 reputation.

CSV / ledger-shaped exports, per-period statements.

Tiered pricing, free-tier quota, per-caller rate limits, promotional zero-price windows.

Accept any SEP-41 asset, receive USDC atomically. Stellar-native; no equivalent on Base or Solana.

Agent pays USDC, merchant receives local fiat in a bank account. The demo no other x402 stack can run — highest-leverage item on this page.

MPP session support alongside x402 exact / upto, one install. MPP is live on Stellar with Stripe/Tempo distribution; merchants shouldn't have to bet on a protocol.

Depends on: MPP session semantics stabilising

Per-call cap, per-day cap, per-merchant allowlist, per-scheme rules. SDF already points x402 builders at these contracts; the gap is the x402-shaped policy set.

Protocol 27 “Zipper” makes authentication delegation for smart contract accounts first-class; revisit once mainnet.

Depends on: Protocol 27 on mainnet

Per-agent attribution, budget burn-down, kill switch.

Every payment an agent made, exportable, verifiable.

Python, LangChain, Vercel AI SDK, OpenAI Agents SDK, plus the MCP path. @x402/fetch alone is not where agents actually live.

Fund a Stellar agent wallet from Base/other USDC in seconds. Circle CCTP launched on Stellar in May; kills the cold-start objection.

Consumer wallet as the human face of the agent budget. First-party; native walras integration.

The catalog as an order book of services. Consumes the same public /discovery endpoints agents use — permanent visible dogfooding.

/discovery/resources facilitator deployment down

Pay a catalogued service in-browser with Freighter for a cent. The fastest possible “I get it” moment.

Derived from real settlement data, not self-reported.

Settlement count, success rate, median latency. Ethereum got ERC-8004 for agent identity in January; Stellar has nothing yet.

Categories, editorially grouped sets, no paid placement.

House listings receive zero ranking preference, verifiable in code; ranking config published. Locked decision — it survives every layer above.

Settled transactions, distinct payers, catalog size, median latency, all networks. Funded as the T3 status page, then extended.

A shareable page per catalogued service.

The same suite run against every Stellar facilitator, results public.

What got paid for on Stellar this month.

Settlement executes inside the Confidential Token wrapper; the client proof shows the hidden transfer pays the quoted price to the advertised recipient; the facilitator verifies a proof instead of a plaintext amount.

Depends on: Confidential Tokens leaving Developer Preview

Against the OpenZeppelin suite + Nethermind UltraHonk verifier.

Depends on: exact-confidential scheme spec

Metered billing where the settled amount is hidden — usage volume stays private. Composes with the Layer 0 upto contract.

Depends on: Layer 0 upto contract

The wrapper's disclosure primitive is the receipt: prove a specific payment to a specific party without exposing the rest.

Auditable privacy, not opaque privacy — the framing SDF is already committed to.

Prove “>N settlements, <X% failure rate” for Bazaar ranking without revealing revenue.

SDF holds a Premier seat on the x402 Foundation Governing Board and will want a flagship reason for it.

Authored, implemented, and coordinated through the x402 TSC rather than kept private.

Role-based x402 guide, contributed upstream.

One worked port, config-level.

Through the x402 TSC, once the primitives are mainnet-eligible.

Depends on: Confidential Tokens leaving Developer Preview

Including the ones competing with walras. That's the point.

The project's own machine-readable surface.

T1 code-complete, submitted ahead of target.

Live on testnet with real TVL; T3 + audit submitted; mainnet targeted Sept/Oct 2026.

In development.