Progress
Phase 1: The Wire
Core x402 payment flow. Agent pays service. Service delivers. Receipt generated.
Core SDK
Wallet management, x402 client, basic receipts
Wallet Providers
Coinbase CDP Server Wallets, raw private key, custom providers
Framework Integrations
OpenClaw skill published, MCP server with auto-detect CDP/raw key
Transaction Logging
On-chain settlement records and local transaction history
Phase 2: The Eyes
Observability. Developers can see what their agents are spending.
Agent hits 80% of daily budget, developer gets notified. At 100%, the agent pauses. No empty wallets at 3am.
RoCS: Return on Cognitive Spend
RoCS: Return on Cognitive Spend
The metric that defines agent commerce performance.A RoCS of 3.0 means the agent creates 1 it spends. Below 1.0, it’s losing money. Works the same whether you’re running one agent or a hundred.The SDK computes RoCS automatically from transaction data. Developers add one line:
Transaction Taxonomy
Transaction Taxonomy
Every transaction is auto-classified by category:
| Category | Examples |
|---|---|
| Inference | LLM API calls, embedding generation |
| Data | Research APIs, enrichment, web scraping |
| Compute | Server provisioning, GPU rental |
| Tooling | SaaS APIs, developer tools |
| Commerce | Physical purchases, domain registration |
| Agent-to-Agent | SoulMarket jobs, peer agent services |
Commerce Dashboard
Commerce Dashboard
Developer View — Real-time spend rate, RoCS trend, success/failure rates, cost breakdown by category, budget utilization, searchable receipt log.Fleet View — All agents ranked by RoCS, aggregate spend, anomaly detection, cost allocation by agent, fleet-level P&L.Ecosystem View (public) — Total transaction volume, active agents, average RoCS, growth metrics.
Budgets and Alerts
Budgets and Alerts
Phase 3: The Proof
Verification. Every transaction becomes a provable event.
Cryptographic Receipts
Cryptographic Receipts
Every x402 transaction produces a signed receipt:Receipts are signed by OneShot, reference on-chain settlement, and are machine-readable.
Failure Mode Handling
Failure Mode Handling
| Failure | SDK Response |
|---|---|
| Payment accepted, no delivery | Auto-dispute, receipt flagged, refund requested |
| Price changed between quote and pay | Reject, re-quote, log discrepancy |
| Double-charge | Block second payment, alert developer |
| Service unreachable after payment | Receipt marked incomplete, auto-retry with backoff |
| Repeated small charges (drain) | Pause agent, alert developer, block service |
| Infinite loop (A pays B pays A) | Cycle detection, break loop, alert both parties |
| Settlement failure | Queue retry, hold service delivery |
Circuit Breakers
Circuit Breakers
Built-in per-service circuit breakers:Service fails 3 times in a minute, the SDK stops paying it. Automatically.
Phase 4: The Score
Public benchmarks. OneShot becomes the standard.
Public Benchmark
Public Benchmark
| Metric | Ecosystem Avg | Top 10% | Your Agent |
|---|---|---|---|
| RoCS | 2.4x | 5.1x | 3.8x |
| Transaction Success Rate | 94.2% | 99.1% | 97.3% |
| Avg Settlement Time | 1.2s | 0.4s | 0.8s |
| Cost per Transaction | $0.003 | $0.001 | $0.002 |
Agent Autonomy Gradient
Agent Autonomy Gradient
The SDK tracks agent maturity automatically:
Each graduation is automatic (SDK detects the threshold), verifiable (backed by receipt data), and public (opt-in via SoulMarket).
| Level | Description | Trigger |
|---|---|---|
| 0 | Agent runs on human’s infra | Wallet created |
| 1 | Agent has wallet, can receive | First incoming payment |
| 2 | Agent earns revenue via x402 | 10+ paid transactions as seller |
| 3 | Agent covers its own costs | 30-day rolling RoCS above 1.0 |
| 4 | Agent purchases own infrastructure | Programmatic compute/infra purchase |
| 5 | Agent replicates to owned infra | Agent deploys new instance with own funds |
Framework Integrations
| Framework | Integration | Status |
|---|---|---|
| OpenClaw | SKILL.md on ClawHub | Live |
| MCP Server | @oneshot-agent/mcp-server | Live |
| LangChain | langchain-oneshot | Live |
| Virtuals GAME SDK | game-plugin-oneshot | Live |
| CrewAI | OneShotTools.crewai() | Planned |
| AutoGen | OneShotTools.autogen() | Planned |
| OpenAI Agents SDK | OneShotTools.openai() | Planned |
Priority Overview
| Feature | Impact | Phase |
|---|---|---|
| Core x402 flow | Table stakes | 1 |
| Wallet provider abstraction | Security & adoption | 1 |
| OpenClaw + MCP integration | Distribution | 1 |
| RoCS metric | Targeting system | 2 |
| Transaction taxonomy | Observability foundation | 2 |
| Developer dashboard | Retention and trust | 2 |
| Budget alerts | Safety | 2 |
| Cryptographic receipts | Verification | 3 |
| Failure mode handling | Tail-risk safety | 3 |
| Circuit breakers | Automatic protection | 3 |
| Public benchmark | Industry standard | 4 |
| Autonomy gradient | Agent maturity tracking | 4 |
| Fleet dashboard | Enterprise retention | 5 |
| Red-team report | Trust and marketing | 5 |