How POLYROB compares
POLYROB is a goal-achieving, self-evolving, crypto-native agent — and a base you can ship products on. It's MIT-licensed and self-hosted: the full engine, free forever, yours to fork and run — not a limited open core wrapped around a paid cloud. The table below compares capabilities and architecture — the axes we can state about ourselves first-party.
| Capability | POLYROB | Hermes | OpenClaw |
|---|---|---|---|
| License | MIT | MIT | MIT |
| Goal-achieving autonomy | ✓ durable goal board + cron | ✓ kanban + DAG board | ✗ |
| Proactive self-wake | ✓ self-wake + change-gate | ✗ | ✗ |
| Self-evolving skills | ✓ writes + curates | ✓ learning loop | ✓ workspace |
| Install third-party skills | ✓ local / GitHub / URL, scanned + quarantined | ⚠ bundled | ⚠ bundled |
| Cross-session memory | ✓ H-MEM hierarchical | ✓ FTS5 + pluggable | ⚠ session / workspace files |
| Agent wallet | ✓ policy-capped, audited — every write simulated and its deltas asserted before signing | ✗ | ✗ |
| On-chain payments (x402) | ✓ Base USDC (Base Sepolia for testnet); optional direct-settlement USDC on Solana | ✗ | ✗ |
| Crypto integrations | ✓ guarded EVM + Solana swaps/bridges; separate Hyperliquid and Polymarket tools (dry-run by default) | ✗ | ✗ |
| On-chain identity + reputation (ERC-8004) | ✓ 3 registries | ✗ | ✗ |
| Security model | ✓ four access roles + capability gates | ⚠ pairing | ✓ sandbox |
| Prompt-injection defense | ✓ untrusted-output wrapping | ⚠ partial | ⚠ partial |
| Multi-tenant + billing | ✓ isolation + metering + credits | ✗ single-user | ✗ single-user |
| Standard agent APIs | ✓ REST + OpenAI-compat + A2A | ✗ | ✗ |
| Provider choice | ✓ built-ins + OAuth plans + declared endpoints | ✓ | ✓ |
| MCP support | ✓ client + server | ✓ client + server | ⚠ client only |
| Browser automation | ✓ Playwright | ✓ | ✓ |
| Chat surfaces (one contract) | ✓ Telegram / WhatsApp / email / Discord / Slack / Signal / X | ✓ more channels | ✓ more channels |
✓ supported ⚠ partial/limited ✗ not supported
When to choose POLYROB
You want goals chased, not chatted
Durable goals + autonomy loops that run unattended, survive restarts, and deliver — and an agent that evolves its own skills as it goes.
You want a crypto-native agent
A built-in wallet and x402 pay-per-request in Base USDC (Base Sepolia on testnet), with optional direct-settlement USDC invoices on Solana — the agent can pay for HTTP resources and charge for its own work. Those payment rails are separate from opt-in DeFi: guarded swaps on Ethereum, Base, Arbitrum, Polygon, Robinhood Chain and Solana; bridging; fixed-supply token deployment; Pons V2 launchpad actions; and EVM dapp access. Hyperliquid and Polymarket are separate venue tools: live orders are off and dry-run by default.
You're building an agent product
Multi-tenant isolation, metering + credits, and standard APIs (REST / OpenAI-compat / A2A) are in the core — ship a paid agent without a backend. MIT, self-hosted.