Deploy multi-step operators
Run agents that research, plan, call tools, and stream progress back to product UIs.
Open this page in Chrome or Brave with MetaMask, Rabby, or Coinbase Wallet installed. The in-app preview browser usually has no wallet extension.
Protector runs autonomous AI workflows with wallet identity, signed skill manifests, and live safety checks before any action reaches production.
The concept keeps the Web3 agent-infrastructure shape from the reference, but the name, copy, colors, and visual system are new.
Compose wallet-bound agents, signed skills, streaming responses, and safety detectors from one production console.
Run agents that research, plan, call tools, and stream progress back to product UIs.
Wallet sessions unlock the network while fair-use rules keep throughput stable.
Ship markdown skill files, bind them to tools, and publish signed manifests.
Deterministic runtime keys are derived from signed wallet sessions.
Send token deltas, tool updates, and completion events over JSON streams.
Route across model providers with policies for latency, cost, and tool reliability.
Every skill and prompt bundle is checked for secrets, injection, risky commands, and network exfiltration before it can run.
Finds private keys, API tokens, session cookies, JWTs, and mnemonic-style payloads before publish.
Flags role-swap, indirect injection, hidden instructions, and hostile tool-output chains.
Blocks reverse shells, destructive cleanup, environment scraping, and suspicious command recipes.
Detects path traversal, webhook drops, DNS tricks, and outbound transfer patterns.
Compares declared permissions against the files, networks, and tools the skill actually touches.
Checks URLs against phishing feeds, typosquat rules, and fresh malicious infrastructure signals.
From wallet signature to first agent response in less than a minute.
Sign in with an EVM wallet. The address becomes your identity across the network.
Pick a public template or publish a signed private skill manifest for your app.
Receive token deltas, tool traces, and final results through one event stream.
A concept utility token for attribution, governance, contributor rewards, and staking skill reputation.
Skill authors and node operators earn reputation-weighted rewards based on usage.
Stake against published skills to signal trust and share upside from adoption.
Vote on routing policy, sentinel rules, and new chain deployments.
A lightweight docs area is now built into the page so the Docs buttons lead somewhere useful.
Users authenticate with an injected EVM wallet. The public wallet address becomes the runtime identity.
await window.ethereum.request({ method: "eth_requestAccounts" })
Publish a signed skill manifest that defines the tools, permissions, and model routing policy.
protector skill publish ./agent.md --network base
Send agent events to your app with token deltas, tool traces, safety checks, and final output.
GET /v1/agents/protector/events
Launch a wallet-native agent runtime with streaming responses, signed skill manifests, and safety scans built into the path to production.
Placeholder launch copy for the rebrand concept. Swap these answers for production legal and token details before going live.
A wallet-native runtime for deploying AI agents with signed skills, streaming responses, and pre-execution security scans.
No. A wallet signature creates the runtime session, and the same address can recover the same identity.
Secrets, prompt injection, risky commands, data exfiltration, permission drift, suspicious URLs, and unsafe tool boundaries.
This concept positions Base as the primary network with Ethereum identity support and additional L2s later.
Yes. Teams can publish signed manifests that route to their own model or tool endpoints.
The concept treats the token as governance and contributor infrastructure, not as a hard gate for basic runtime access.