cred402/ROADMAP.md
caelum0x 5cf86893e3 feat(contracts): deploy Cred402 Odra contracts to Casper Testnet (Casper 2.0)
Migrate the contract suite to Odra 2.x (builds on the Casper 2.0 ABI; fixes the
1.x nightly/edition2024 toolchain deadlock) and deploy 9 contracts live to
casper-test via the Odra livenet deployer:

  AgentRegistry · X402ReceiptRegistry · RWAEvidenceRegistry · RWAAssetRegistry
  ReputationEngine · AgentCreditPool · RiskPolicyManager · AgentPassport · DisputeCourt

- contracts: odra "2", no_std headers, build.rs (odra-build), per-crate Odra.toml +
  build_contract/livenet bins; pinned nightly-2025-01-15 for reproducible wasm.
- tooling: scaffold_livenet.py, deploy_all_testnet.sh, deploy_one.mjs.
- deploys.testnet.json: live manifest; README: on-chain contract hashes.
- Deployer key kept out of git (.secrets/ gitignored); wasm artifacts gitignored.

Verified on-chain: AgentRegistry ContractPackage present in global state.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-17 02:34:49 +03:00

34 KiB
Raw Permalink Blame History

Cred402 Roadmap

This repo implements the protocol heart of the Cred402 production blueprint (distilled in PRODUCTION.md) as working, tested, runnable code, and maps the remaining production-operations scaffolding as roadmap. Honesty about what is and isn't built matters more than empty stub files.

✅ Built and verified (this repo)

Blueprint area Status
Agent Passport (Product A) ✅ lib/ledger/contracts/agent_passport.ts, /api/passport/:id, dashboard
x402 Receipt Network (Product B) ✅ real 402→sign→200 flow, on-chain receipt commitments
RWA Evidence Graph (Product C) ✅ evidence registry + asset registry, linked to receipts
Agent Credit Pool (Product D) ✅ deposit/open/draw/repay/liquidate, fees, health factor
Dispute / Slashing / Governance (Product E) ✅ DisputeCourt + SlashingVault + Governance + DisputeJudgeAgent
Reputation Engine (§6.6) ✅ multi-dimensional, on-chain composite score
Risk Policy Manager (§6.8) ✅ upgradable v1→v2, governance-bound
Agent system (§8) ✅ Buyer, Seller, Credit, Treasury, Watchdog, DisputeJudge, LiquidityRouter
MCP server (§12) ✅ 16 tools + 6 resources over stdio JSON-RPC (mcp/)
TypeScript SDK (§4) ✅ sdk/ Cred402Client (REST + x402 + SSE)
API Gateway (REST + SSE + x402) (§7.1–7.2) ✅ api/ zero-dep node:http
Console / dashboard (§3) ✅ Vite + React, 6 tabs, live event feed
Casper contracts (Odra/Rust) (§6) ✅ 14 Odra crates under contracts/ (registry, passport, x402 receipts, RWA asset + evidence, reputation, credit pool, risk policy, dispute court, slashing vault, governance + p6 fiat-receipt / operator-verification / realfi-attestation) — all compile; 11 unit-tested
Event model (§13) ✅ streaming events → SSE → watchdog + dashboard
Tests ✅ 45 node:test cases (test/) + 39 Odra contract tests across all 11 Casper crates, run in the OdraVM via cd contracts && cargo test (no WASM/testnet needed)

✅ p3 omnichain layer (Casper-rooted, chain-executed) — built

Blueprint area Status
Cross-chain standards (CAID/ABE/URE/UAID/EAE/CAN) ✅ crosschain/standards — real dual-sig (ed25519+secp256k1), blake2b ids, JSON schemas + validator
AddressBindingRegistry ✅ dual-signature-verified bindings (lib/ledger/contracts/address_binding_registry.ts)
ExternalReceiptRegistry ✅ anchors non-Casper x402 receipts; reputation settles on Casper
GlobalExposureManager ✅ per-agent cap across all chains — the over-borrow guard
CreditAuthorizationNotes ✅ short-lived Casper-policy-signed CANs; reserve exposure on issue
UpgradeManager ✅ contract versions + upgrade history
ChainAdapter SDK ✅ packages/chain-adapters — CasperAdapter (root) + EvmAdapter/EvmSatelliteVault + CosmosAdapter/CosmosSatelliteVault (CosmWasm, sha256 tx ids)
End-to-end omnichain flow ✅ npm run demo:multichain + 5 p3 tests — two satellite families (Base + Osmosis) sharing one Casper-rooted exposure cap (bind → earn → anchor → CAN → lend → repay)
Satellite contracts (EVM/Solana/Cosmos/Move/Bitcoin) ✅ contracts/{evm,solana,cosmos,move,bitcoin}
Casper-root relayer + proof service ✅ crosschain/relayers (event→proof→anchor) + crosschain/proof-service (signed blake2b Merkle batch + inclusion proofs). Wired into npm run demo:multichain and tested in test/relayer.test.ts
Real RWA data ✅ Open-Meteo solar + PV physics model (api/rwa_data) — no mock
Real crypto ✅ @noble/curves secp256k1, @noble/hashes keccak256/blake2b

Phases 1–7 of the p3 expansion sequence are represented: Casper core (p1/p2), address bindings, external receipt anchoring, evidence mirrors, Casper-issued credit notes, satellite credit vaults, and the global agent credit market.

✅ p4 fraud & attack hardening (§13) — built

Attack (p4 §13) Mitigation status
1 — wash receipts ✅ FraudService: reciprocal-loop + operator-linkage graph detection, revenue concentration, velocity, and off-market pricing-band anomaly (lib/services/fraud_service.ts)
2 — Sybil agent swarm ✅ operator identity graph — flags one operator controlling ≥3 agents
5 — credit pool drain ✅ AgentCreditPool exposure caps + freeze/liquidate + GlobalExposureManager
6 — cross-chain double borrow ✅ Casper-issued CAN + exposure reservation before any satellite draw
7 — relayer lies ✅ ProofService signed Merkle batch + inclusion proofs; relayer keys allowlisted before anchor

The fraud score gates underwriting (econ.credit.underwrite refuses high-fraud agents); tested in test/protocol.test.ts.

✅ p4 economics/marketplace/trust-ladder + p5 reason codes — built

Blueprint area Status
Marketplace (p4 §18) ✅ lib/services/marketplace.ts — full 16-category taxonomy, 7 pricing strategies (fixed/dynamic/auction/subscription/reputation-tiered/urgency/data-cost-plus), trust-enriched listings ranked by reputation
Protocol economics / fee model (p4 §11) ✅ lib/core/economics.ts — facilitator/origination/late fees, 10% interest spread, 50/25/25 slash route, honest LP poolHealth (realized APY + risk flags, no fake APY)
Trust ladder Stage 2 (p4 §26) ✅ crosschain/trust-ladder/multi_relayer.ts — bonded multi-relayer quorum + challenge window + fraud-proof bond slashing
Trust ladder Stage 3 ✅ crosschain/trust-ladder/finality.ts — per-chain confirmation + time finality policy
Trust ladder Stage 4 ✅ crosschain/trust-ladder/proof_types.ts — pluggable proof-type registry (merkle + threshold real; light_client/zk honestly rejected until built)
Credit reason codes (p5 §15) ✅ lib/core/reason_codes.ts — 14 structured positive/negative codes derived from on-chain signals, attached to every credit decision and shown in npm run demo
Demo ✅ npm run demo:marketplace (marketplace + economics + trust ladder); reason codes in npm run demo

✅ p6 RealFi Bridge — built

Blueprint area Status
RealFi envelopes (FRE/OVE/BVE) ✅ lib/realfi/envelopes.ts — privacy-preserving (hashes only, PII never on-chain), blake2b commitments + validators
FiatReceiptRegistry (p6) ✅ lib/ledger/contracts/fiat_receipt_registry.ts + Odra contracts/fiat_receipt_registry — Stripe-equivalent of the x402 receipt, provider-event idempotency
OperatorVerificationRegistry ✅ TS + Odra crate — Stripe-Identity-style operator KYB attestations
RealFiAttestationRegistry ✅ TS + Odra crate — generic bank/cashflow/chargeback/sanctions attestations
RealFi Bridge service ✅ lib/services/realfi_bridge.ts — turns Stripe/Plaid payloads into hashed on-chain envelopes
RealFi credit scoring (p6 §864) ✅ lib/services/realfi_credit.ts — bounded ±uplift (operator/fiat/bank/chargeback), capped so fiat never dominates Casper-native receipts; wired into CreditAgent + reason codes
Surfaces ✅ wired end-to-end: REST (GET /api/realfi, POST /api/demo/realfi, /api/realfi/verify-operator/fiat-receipt/chargeback), 3 MCP tools (19 total), console RealFi tab, and npm run demo:realfi (anonymous → verified +uplift → chargeback cut, zero PII on-chain)
EVM RealFi mirror (Solidity) ✅ contracts/evm/src/Cred402RealFiMirror.sol — mirrors fiat receipt + operator-verification commitments on EVM (emits for the Casper-root relayer); wired into Deploy.s.sol; compiles under Foundry (via_ir now default)

✅ p7 §9/§11 — built

  • THIRD_PARTY.md (§9) — honest open-source attribution + originality boundary.
  • npm run ship:check (§11) — scripts/ship_first_check.ts reports the ship-first stack status against the repo (7 integrated · 4 own-impl · 1 transport-swap pending).

✅ Console surfacing (p4 §11/§18, p5 §15)

  • Credit Pool tab — click a credit line to see its structured reason codes (/api/credit/explain/:id) + a Protocol Economics panel (honest realized APY, fees, loss rate, risk flags via /api/economics).
  • Marketplace tab — trust-ranked agent service listings (/api/marketplace).

THIRD_PARTY.md (p7 §9) records the open-source rails used vs. the protocol Cred402 owns.

✅ Production API gateway (p2 §7.1) — built

The hackathon server is now a real service. lib/gateway/ + the versioned /v1 API (api/v1/router.ts) add the production concerns the console-facing /api/* routes lacked:

Concern Status
Config + fail-fast secrets ✅ config.ts — typed env config; refuses to boot on mainnet/auth without required secrets
Structured logging ✅ logger.ts — JSON logs, level gate, secret redaction, request-scoped child loggers
Error envelope ✅ errors.ts — typed ApiErrors → stable codes/status; one {success,data|error,request_id} shape
Input validation ✅ validation.ts — composable, dependency-free schema parser at every boundary (no blind as)
API keys + scopes ✅ api_keys.ts — hashed storage, constant-time verify, read/write/admin scopes
Rate limiting ✅ rate_limit.ts — per-identity token bucket
Idempotency ✅ idempotency.ts — Idempotency-Key replay protection on mutations
Webhooks ✅ webhooks.ts — HMAC-signed delivery + retry/backoff, fanned from the event bus
Durable persistence ✅ persistence.ts — append-only NDJSON event journal (CRED402_DATA_DIR) + atomic snapshots; replaces in-memory-only

/v1 routes for agents, receipts, credit (pool/lines/draw/repay/explain), economics, marketplace, disputes, realfi, and admin (api-keys, webhooks) — each authenticated, rate-limited, validated, and enveloped. The console keeps the unversioned /api/* routes.

✅ Production services & infra — built

Area Status
Compliance service (§7.9) ✅ lib/compliance/ — sanctions screening (subject + OFAC-style jurisdiction lists), jurisdiction lending policy, KYB gate, data-retention policy; wired into CreditAgent.underwrite/explain (refuses sanctioned operators) + /v1/compliance/agents/:id
Go event indexer (§7.4) ✅ services/event-indexer/ — replays the NDJSON journal into agent/pool/dispute projections with a seq checkpoint; go build/go vet clean; verified against a live journal. sink.Sink is the Postgres seam
Casper transport (live Testnet) ✅ lib/casper/ — real fetch JSON-RPC client (reads work with no SDK), typed deploy/runtime-arg construction + casper-client renderer, CasperTransport interface with casper-js-sdk reduced to an injected DeploySigner. npm run casper:health
Infra ✅ infra/ — multi-stage Dockerfile (non-root, healthchecked), docker-compose.yml (api + indexer + postgres), Helm chart (helm lint clean: deployment/service/ingress/hpa/pvc/secret-ref), cloud-agnostic Terraform module

✅ Production-completion layer — built

Area Status
Production plan ✅ PRODUCTION.md — full-product plan: surfaces, protocol core, services, security, deployment, launch sequence, doc→code map
OpenAPI 3.1 ✅ packages/openapi/cred402.v1.yaml — all 27 /v1 operations, envelope + auth + schemas; redocly lint clean
GraphQL API (p2 §7.1) ✅ lib/graphql/ — typed read surface (graphql ref impl) at POST/GET /graphql; agents, pool, analytics, marketplace, search, compliance
Python risk-engine (p2 §7.6/§7.8) ✅ services/risk-engine/ — logistic PD credit model + linear (SHAP-like) reason attribution + Tarjan-SCC fraud graph; HTTP API + CLI, runs live
Casper transport (live seam) ✅ lib/casper/ — LedgerTransport (real, executes entry points) + Ed25519DeploySigner (real blake2b hash + ed25519 signature + account_put_deploy) + JSON-RPC reads. Two working CasperTransport impls
4-language SDKs/CLIs ✅ cli/ (TS), sdk/python, sdk/go, sdk/rust — all build + run live
New product APIs/pages ✅ analytics/notifications/search services + Analytics & Explorer console pages + notification bell

✅ Product expansion — built

Area Status
Developer portal (console) ✅ frontend/src/components/Developer.tsx — mint scoped API keys, subscribe webhooks (one-time secrets), live GraphQL console
Agent 360° drill-down ✅ /api/agent-profile/:id + /v1/agents/:id/profile + AgentDetail.tsx overlay (passport, credit reason codes, compliance, RealFi, receipts, reputation history); click any leaderboard row
Prometheus /metrics ✅ lib/gateway/metrics.ts — pool/agents/disputes/defaults gauges + per-event counters in exposition format
@cred402/sdk (TS) ✅ packages/ts-sdk/ — full typed client (every /v1 route + GraphQL + idempotency + webhook signature verify)
Python agent orchestrator ✅ services/agent-orchestrator/ — policy engine OUTSIDE the LLM (spending limits, tool permissions, circuit breaker, approval gate), planner, audit log, pure-stdlib Ed25519 for x402; demonstrably blocks an over-limit draw
Notification service ✅ services/notification-service/ — multi-channel (console/webhook/Slack/Discord), templates, dedupe + retry, polls the live feed
Webhook receiver example ✅ examples/webhook-receiver/ — verifies HMAC signatures via the SDK; full subscribe→deliver→verify loop confirmed (27/27 verified)

✅ Bureau & gateway depth — built

Area Status
Credit report (headline bureau artifact) ✅ lib/services/credit_report.ts — FICO-style report (score+band, PD estimate, recommended terms, +/− factors, payment history, public records, inquiries, revenue, compliance); /v1/agents/:id/credit-report + /api/credit-report/:id + console CreditReportView (gauge) in the agent overlay
GraphQL mutations ✅ Mutation type — registerAgent/openCreditLine/draw/repay/verifyOperator/openDispute; SDL exported via npm run graphql:sdl → packages/graphql/schema.graphql
Ops / incident board ✅ /api/incidents + console Ops tab — fraud watchlist, frozen/defaulted lines, open disputes, emergency pause switches
x402 playground ✅ /api/x402/buy + console x402 tab — runs the real 402→sign→200 flow and visualizes challenge headers + receipt + signed report
Prometheus HTTP metrics ✅ /metrics now includes cred402_http_requests_total{route,status} from the v1 gateway
Pagination + filtering ✅ /v1/agents (service_type, min_reputation) and /v1/receipts (status, seller) support ?limit/offset — array shape preserved (SDK-compatible)

✅ Functional commerce & live data — built

Area Status
Agent-to-agent marketplace commerce ✅ state.marketplacePurchase — any agent buys a listed service; records a real facilitator-settled x402 receipt that builds the seller's revenue + reputation (the flywheel). REST (/api/marketplace/purchase, /v1/marketplace/purchase), console Buy button, GraphQL purchaseListing mutation
SSE live analytics ✅ /api/analytics/stream pushes the analytics snapshot on each event (throttled); the Analytics page consumes it — live, no polling
Batch underwriting ✅ /v1/credit/underwrite-batch — underwrite up to 100 agents in one call with per-item ok/error
CSV export, simulation, status doctor, public HTML report ✅ /api/export/*.csv, npm run simulate, npm run status, GET /report/:id

✅ Full economic loop driveable from every surface — built

Every step of the protocol's economic cycle is now a real API action (REST /api

  • /v1, GraphQL mutations, console controls):
Step Endpoint(s)
LP deposit / withdraw /v1/credit/deposit, /v1/credit/withdraw (free-liquidity guard) + console buttons
Agent register / stake /v1/agents, /v1/agents/:id/stake (+stakeAgent GraphQL) — stake boosts the credit line
Earn (x402 / marketplace) /api/x402/buy, /api/marketplace/purchase (+purchaseListing GraphQL)
Underwrite (single / batch) /v1/credit/lines, /v1/credit/underwrite-batch
Draw / repay /v1/credit/lines/:id/draw, /repay
Dispute open / resolve / slash /v1/disputes, /v1/disputes/:id/verdict (slashes + reputation hit) + console resolve buttons

Integrator artifacts: packages/openapi/cred402.v1.yaml (OpenAPI 3.1), cred402.http (REST Client), cred402.postman_collection.json, packages/graphql/schema.graphql (SDL). Live surfaces: /graphql (+/graphiql +/graphql/stream live queries), /api/analytics/stream (SSE), /metrics (Prometheus incl. HTTP request counts), /report/:id (public HTML credit report).

✅ DeFi pool, insurance, lifecycle mechanics — built

Mechanic Endpoint(s)
LP positions + pro-rata yield /api/lp, /v1/credit/lp, GraphQL lp — console LP positions table
LP deposit / withdraw /v1/credit/deposit · /withdraw (free-liquidity guard)
Agent self-listing /v1/marketplace/listings, GraphQL createListing — console "List a service" form
Insurance claims /v1/insurance/claim — pays from the slashing insurance reserve (dispute slash → 30% reserve → LP claim; reserve guard)
Reputation time-decay /v1/reputation/decay (p2 §6.6) — inactive agents drift to a floor
Credit-line health + freeze /v1/credit/health, /v1/credit/lines/:id/freeze

The extended npm run simulate exercises the entire loop (deposits, staking, agent-to-agent trading, dispute→verdict→slash→insurance reserve). 39 /v1 routes, 14 console tabs, GraphQL read+write parity.

✅ Discovery, trust & multichain depth — built

Area Status
Agent discovery / ranking ✅ lib/services/discovery.ts — composite buyer-facing score fusing reputation (0.35) + credit (0.25) + web-of-trust (0.20) + tier bonus − fraud penalty, clamped 0..100; filter by service_type/min_reputation/min_score. /v1/discovery + console Discovery tab (3 tests)
Web of trust (attestations) ✅ lib/services/attestation_graph.ts — agents vouch for each other with anti-Sybil caps (attester rep ≥ 60, ≤ 6 boost/target, one vouch/pair); /v1/attestations (POST) + /v1/attestations/graph + per-agent; console Trust tab renders the directed graph as SVG
Per-jurisdiction compliance report ✅ lib/services/compliance_report.ts — operators grouped by jurisdiction with KYB coverage + sanctions exposure; /v1/compliance/report + console Compliance tab
Reputation tiers ✅ lib/services/reputation_tiers.ts — bronze→diamond badges with credit-multiplier + origination-discount perks, applied in underwriting; /v1/agents/:id/tier, /v1/tiers
Cosmos + Solana satellites ✅ packages/chain-adapters/.../{cosmos,solana} — CosmosAdapter/CosmosSatelliteVault (CosmWasm, sha256) and SolanaAdapter/SolanaSatelliteVault (Anchor/SPL, base58 sigs); npm run demo:multichain runs three satellite families (Base + Osmosis + Solana) under one shared Casper-rooted exposure cap

| Credit what-if simulator | ✅ lib/services/credit_simulator.ts — read-only underwriting preview: runs the live risk policy + governance cap against hypothetical signals without mutating state. POST /v1/credit/simulate + console Risk-tab panel (4 tests) | | Peer benchmark | ✅ lib/services/peer_benchmark.ts — percentile + rank of an agent vs its service-type cohort (reputation/credit/revenue/fraud). /v1/agents/:id/benchmark + AgentDetail overlay section (3 tests) | | Satellite families: Solana + Move | ✅ SolanaAdapter/SolanaSatelliteVault (Anchor/SPL, base58 sigs) and MoveAdapter/MoveSatelliteVault (Aptos/Sui Move, 0x-hex tx hashes) — four satellite families (EVM/Cosmos/Solana/Move) execute credit against Casper CANs under one shared exposure cap; demo:multichain runs Base + Osmosis + Solana |

| Credit pre-approval offers | ✅ lib/services/credit_offers.ts — time-bounded offer (terms from the live underwriter) → agent accepts to open a line at locked terms, or it expires. /v1/credit/offers (issue/list/accept/decline) + console Credit Pool card + MCP tools + cred402 credit offer/offers/accept/decline CLI (6 tests) | | CLI bureau commands | ✅ cred402 credit simulate (what-if preview) + the offer lifecycle subcommands — the TS CLI reaches the new /v1 surfaces with human + --json output | | Agent credit history (credit file) | ✅ lib/services/credit_history.ts — every on-chain event concerning an agent, chronological + categorized (identity/revenue/credit/dispute/reputation/crosschain), built from the canonical event log. /v1/agents/:id/history + AgentDetail "History" view + MCP tool (2 tests) | | Python SDK parity | ✅ sdk/python — discovery.search/attest/attestation_graph, credit.portfolio/simulate/offers/issue_offer/accept_offer/decline_offer, agents.benchmark/history reach the new bureau surfaces; integrator .http collection extended to match | | GraphQL parity | ✅ lib/graphql — discovery/portfolio/attestationGraph/benchmark/creditHistory/scoreTrend/readiness/riskAlerts/yieldProjection/fleetOverview/simulateCredit/creditOffers queries + issueCreditOffer/acceptCreditOffer mutations; every bureau read surface now resolves over GraphQL, SDL re-exported to packages/graphql/schema.graphql | | Go + Rust SDK parity | ✅ sdk/go (Discover/Portfolio/AttestationGraph/Benchmark/CreditHistory/SimulateCredit/CreditOffers/IssueCreditOffer/AcceptCreditOffer) and sdk/rust (discover/attestation_graph/attest/portfolio/benchmark/credit_history/simulate_credit/credit_offers/issue_credit_offer/accept_credit_offer) reach the new surfaces — verified live; bureau features now have parity across REST, GraphQL, MCP, CLI, and all four SDKs | | Risk monitoring alerts | ✅ lib/services/risk_alerts.ts — always-on sweep emitting severity-ranked alerts (concentration breach, overdue lines, fraud exposure on drawn credit, frozen/defaulted lines, liquidity stress). /v1/risk/alerts + console Ops-tab card + MCP tool (4 tests) | | @cred402/sdk parity | ✅ packages/ts-sdk — discover/attestationGraph/attest/portfolio/riskAlerts/yieldProjection/benchmark/creditHistory/simulateCredit/creditOffers/issueCreditOffer/acceptCreditOffer/declineCreditOffer; the new bureau surfaces now reach all five SDKs (ts-sdk, Python, Go, Rust, CLI) plus REST/GraphQL/MCP — verified live | | LP deposit preview | ✅ lib/services/lp_deposit_preview.ts — the lender's counterpart to the credit simulator: previews resulting pool share, utilization, and projected annual yield for a hypothetical deposit (no funds moved). /v1/credit/lp-preview?deposit_cspr= + console Credit Pool preview widget + MCP tool + ts-sdk (4 tests) | | LP yield projection | ✅ lib/services/yield_projection.ts — forward LP yield over 30/90/365 days (gross interest, LP share after protocol spread, expected loss, projected net APY); explicitly assumption-driven and kept separate from realized APY. /v1/credit/yield-projection + console Credit Pool card + MCP tool (3 tests) | | Onboarding readiness scorecard | ✅ lib/services/onboarding_scorecard.ts — pass/fail checklist of every credit gate (reputation, disputes, fraud, compliance, revenue, stake, KYB) with guidance + overall readiness %; mirrors the live underwriter gates. /v1/agents/:id/readiness + console Onboard-tab check + MCP tool + ts-sdk (3 tests) | | Offer lifecycle events | ✅ CreditOfferIssued/Accepted/Declined now emit on the canonical event bus → fan out to HMAC webhooks, the SSE stream, and the agent's credit history automatically (2 tests) | | Bureau integration capstone | ✅ test/bureau_integration.test.ts — one end-to-end test drives offer → accept → draw and asserts history, portfolio, risk alerts, yield projection, readiness and discovery all reflect the same shared ledger | | Credit-score & reputation trend | ✅ lib/services/score_trend.ts — reconstructs an agent's score/reputation trajectory (current, net change, points) from the canonical event log. /v1/agents/:id/score-trend + AgentDetail sparklines + MCP tool + ts-sdk (3 tests) | | bureau CLI command group | ✅ cli/commands/bureau.ts — discover/portfolio/alerts/yield/benchmark/readiness/trend/history give ops a one-command view of every analytics surface (human + --json) | | Enriched public credit report | ✅ /report/:id now server-renders trend sparklines (inline SVG, no JS), peer-standing percentile, and credit-readiness alongside the score gauge and reason factors — a complete shareable credit file | | Bureau roster CSV export | ✅ /api/export/bureau.csv — analyst-ready roster joining discovery ranking with per-agent credit readiness; download link on the console Discovery tab | | Bulk agent registration | ✅ POST /v1/agents/batch — register up to 100 agents in one call (operator fleet onboarding) with per-item ok/error so a duplicate doesn't abort the batch + ts-sdk (2 tests) | | Operator fleet overview | ✅ lib/services/fleet_overview.ts — one-call dashboard joining discovery standing + credit readiness + current line for a list of agents (unknown ids flagged). POST /v1/operators/fleet-overview + console Ops-tab fleet check + MCP tool + ts-sdk (2 tests) | | Credit-line review (ratchet-up) | ✅ lib/services/credit_review.ts — periodic re-underwriting that raises a limit when metrics improve, holds otherwise, and never auto-reduces extended credit (reductions stay an explicit freeze). POST /v1/credit/lines/:id/review + GraphQL mutation + console Credit Pool "Review" button + MCP tool + ts-sdk (4 tests) | | Per-agent cross-chain summary | ✅ lib/services/agent_multichain.ts — one agent's address bindings, Casper-anchored external receipts, and Credit Authorization Notes per satellite chain, plus its shared global exposure. /v1/agents/:id/multichain + GraphQL + MCP tool + ts-sdk (3 tests) | | Batch line review | ✅ state.reviewAllCreditLines — periodic portfolio maintenance that re-underwrites every active line (ratchet-up only) and summarizes increased/held/ineligible. POST /v1/credit/review-all (admin) + MCP tool + ts-sdk (2 tests) | | Doctor bureau probes | ✅ npm run status now probes discovery / portfolio / risk-alerts / yield-projection / readiness too — 17/17 surfaces healthy end-to-end against a live server | | Service-category analytics | ✅ lib/services/category_analytics.ts — market intelligence rolled up by service type (agent supply, avg reputation/credit, receipts, revenue, top earner). /v1/analytics/categories + GraphQL + MCP tool + ts-sdk (2 tests) | | Marketplace statistics | ✅ lib/services/marketplace_stats.ts — supply-side stats for the service marketplace: listings by category & pricing strategy, price range, average seller reputation, top sellers. /v1/analytics/marketplace + console Marketplace-tab stats card + GraphQL + ts-sdk (2 tests) | | x402 receipt-network stats | ✅ lib/services/x402_stats.ts — Product-B analytics: total volume, settlement-status breakdown, finalization rate, top sellers/payers, per-service volume. /v1/analytics/x402 + console Network tab (20th) + GraphQL + MCP tool + ts-sdk (2 tests) | | Go + Rust + Python SDK refresh | ✅ brought sdk/go, sdk/rust and sdk/python back to parity with the newest surfaces (health/readiness/score-trend/multichain/similar/credit-cost/compare/batch-register on agents; categories/movers/disputes/x402/risk-alerts/yield/config analytics) — all build/import clean and verified live | | Protocol config (rulebook) | ✅ lib/services/protocol_config.ts — self-documenting live parameters: fee schedule, credit gates (min reputation, max exposure), and the full reputation-tier perk table. /v1/config + console Developer-tab rulebook + GraphQL config + MCP tool + ts-sdk (1 test) | | Bureau lifecycle demo | ✅ npm run demo:bureau (scripts/demo_bureau.ts) — a narrated end-to-end walk: readiness → what-if → offer → accept → draw cost → health → benchmark → compare → trend → multichain → protocol analytics. Living documentation + smoke test against a live server | | bureau CLI refresh | ✅ added health/market/x402/disputes/config subcommands so the CLI reaches every analytics surface (human + --json) | | Deploy/dev fixes | ✅ added frontend/src/vite-env.d.ts (types import.meta.env for the deployed cross-origin streamUrl config — unblocks tsc -b/Vercel build) and extended the Vite dev proxy to /v1·/graphql·/report·/metrics so every bureau surface works under npm run dev | | Contracts dependency pin | ✅ pinned zeroize 1.8.1 · zeroize_derive 1.4.2 · indexmap 2.7.1 · getrandom 0.3.3 in contracts/Cargo.lock — crates.io had published edition2024-requiring releases the pinned mid-2024 nightly can't parse; the 14 Odra contracts compile again | | Dispute statistics | ✅ lib/services/dispute_stats.ts — protocol-level dispute intelligence (totals, open/resolved, outcomes by verdict & type, total slashed, resolution + agent-loss rates, most-disputed agent). /v1/analytics/disputes + console Disputes-tab stats card + GraphQL + MCP tool + ts-sdk (2 tests) | | Reputation movers | ✅ lib/services/reputation_movers.ts — biggest net reputation gainers/losers reconstructed from the event log (momentum, not just level). /v1/analytics/reputation-movers + GraphQL + MCP tool + ts-sdk (3 tests) | | Market section (console) | ✅ the Analytics tab now renders the per-category market table and reputation-mover chips alongside the live leaderboard and trends | | Webhook delivery log | ✅ WebhookService keeps a 200-entry ring buffer of delivery attempts (delivered/failed, http status, retries); GET /v1/webhooks/deliveries (admin, filterable by subscription) + ts-sdk — integrators can debug their webhook endpoints (3 tests) | | Safe-draw advisor | ✅ lib/services/safe_draw.ts — the largest additional draw that keeps a line at/above a target health factor (default 1.5×), bounded by line headroom and pool free liquidity, with the binding constraint named. /v1/agents/:id/safe-draw + GraphQL + MCP tool + ts-sdk (4 tests) | | Credit cost calculator | ✅ lib/services/credit_cost.ts — borrower transparency for a specific draw: upfront origination fee + prorated interest over the line's remaining term + total repayment + effective all-in cost, with headroom guard. /v1/agents/:id/credit-cost?draw_cspr= + GraphQL + MCP tool + ts-sdk (4 tests) | | Agent dossier (one-call) | ✅ lib/services/agent_dossier.ts — integrator snapshot bundling tier + health + readiness + benchmark percentile + reputation momentum + credit line + operator in one round-trip (each section reuses the standalone service, so it never drifts). /v1/agents/:id/dossier + GraphQL + MCP tool + ts-sdk (2 tests) | | Similar agents recommender | ✅ lib/services/similar_agents.ts — "you might also consider": same-category peers ranked by closeness in standing (reputation + credit) and overall discovery strength. /v1/agents/:id/similar + AgentDetail "see also" section + GraphQL + MCP tool + ts-sdk (3 tests) | | Agent health badge | ✅ lib/services/agent_health.ts — a glanceable green/amber/red verdict (worst-of reputation, fraud risk, open disputes, credit-line status) with a composite score + driving factors; distinct from credit-readiness gates. /v1/agents/:id/health + AgentDetail header chip + GraphQL + MCP tool + ts-sdk (4 tests) | | Agent comparison | ✅ lib/services/agent_compare.ts — buyer-facing side-by-side of two agents across discovery/reputation/credit/trust/revenue/fraud/dispute with per-metric + overall winners (lower-is-better metrics invert). /v1/agents/compare?a=&b= + console Discovery "Compare" panel + GraphQL + MCP tool + ts-sdk (3 tests) | | Sign-in with Casper Wallet | ✅ real browser-extension integration (frontend/src/lib/casperWallet.ts, hooks/useCasperWallet.ts, WalletButton with Connect → Sign in → ✓) + backend lib/services/wallet_auth.ts — challenge → wallet ed25519 signature → server-side @noble/curves verification → one-time-nonce session. POST /v1/auth/wallet/challenge + /verify, and GET /v1/auth/wallet/agents returns the account's owned agents ("my agents"). No private key ever leaves the extension (6 tests) |

63 /v1 routes, 19 console tabs, 27 MCP tools, 34 OpenAPI paths, 66 tests.

🔭 Remaining (designed, not yet coded here)

  • Go ports of the relayer/proof-service (p3 §relayers) — the relay→prove→anchor path is implemented and tested in TypeScript (crosschain/); Go binaries are packaging.
  • Python risk/fraud ML port (§7.6, §7.8) — deterministic risk policy + graph FraudService are in TS; the ML scoring variant is an advisory upgrade.
  • GraphQL API (§7.1) — REST /v1 + webhooks + API keys are built; GraphQL is an optional projection.
  • Live deploy signer — the casper-js-sdk DeploySigner implementation (serialize + sign) is the one remaining seam to submit to Testnet; reads + deploy construction are done.

Path to mainnet

  1. Deploy the 8 Odra contracts to Testnet; replace the ledger simulation with casper-js-sdk calls behind the same interfaces.
  2. Stand up the Go indexer against CSPR.cloud streaming events.
  3. Add governance timelock + launch multisig.
  4. Add the Python risk/fraud services as advisory inputs to RiskPolicyManager.
  5. Security audit (see the threat model in SECURITY.md) before mainnet.