dumbtv
dumbtv가 ADB로 Fire TV에서 광고·추적·업데이트 알림을 제거, 루트·앱 변경 없이, 무료 오픈소스
AI 에이전트의 결정과 근거를 서명·변조 방지 기록으로 남겨 Solana에 고정하고 브라우저로 누구나 검증합니다
Signed, tamper-evident records of what your AI agent decided and what it relied on. Anchored on Solana, checked by anyone in their own browser.
보안TypeScript · HTMLGitHub @agenttracewoof온라인운영자 미확인
써 보셨어요? 한 번만 눌러 주세요. 로그인은 필요 없어요.
개인 계정의 공개 저장소 소유자는 GitHub로 로그인해 운영자를 확인하고 소개를 직접 관리할 수 있습니다. 관리자 승인은 필요하지 않습니다.
로그인한 뒤 확인 파일을 만들어 이 저장소의 기본 브랜치에 30분 안에 커밋해 주세요. 다른 브랜치나 포크의 파일은 인정하지 않습니다. 파일을 올린 뒤 아래에서 확인하면 관리자 승인 없이 관리할 수 있습니다. 자격을 유지하려면 파일을 남겨 두세요.
로그인이 필요하면 로그인 화면으로 이동합니다. 로그인 후 이곳으로 돌아와 원하는 확인 버튼을 다시 눌러 주세요.
프로젝트 폴더에서 /nomorevibe verify를 실행하면 소유권을 확인하고 소개를 직접 관리할 수 있습니다.
1분 만에 확인하기README 전체를 불러오는 중입니다. 아래는 저장된 발췌입니다.
Cryptographic provenance and audit layer for AI agent decisions. Every time an agent
decides something, the SDK records a signed decision manifest — which data sources,
which model, which intermediate steps — and anchors its root on Solana. Anyone holding
the link can verify that the record has not changed, without trusting AgentTrace and
without the model, the prompts or the API keys ever being exposed.README 전문 보기 ↗Cryptographic provenance and audit layer for AI agent decisions. Every time an agent decides something, the SDK records a signed decision manifest — which data sources, which model, which intermediate steps — and anchors its root on Solana. Anyone holding the link can verify that the record has not changed, without trusting AgentTrace and without the model, the prompts or the API keys ever being exposed.
| Demo page | https://agenttracewoof.github.io/Agent-Trace/ |
| API | https://agenttrace-api-cr1b.onrender.com |
The page at /verify checks an envelope against the chain in your browser, with
no backend of ours in the path — a public Solana node is the only thing it needs.
The decision page does read our API, and says so when it cannot.
Everything runs on devnet until milestone M5, and devnet is reset periodically: no anchor there is meant to outlive a demo. The API runs on a free instance that spins down when idle, so the first request after a quiet spell can take up to a minute.
It does not run agents, hold funds, or judge whether a decision was good. It proves the decision relied on what it claims to have relied on. Agent identity and reputation are a different layer; so are spending controls.
apps/api ingest, public read, dashboard API
apps/publisher worker: pending decisions → anchor transaction
apps/web dashboard + public decision page
apps/landing static home page, the root of the Pages site
packages/manifest format, hashing, signing, verification — pure, isomorphic
packages/sdk the client SDK agents integrate
packages/verify CLI that verifies from public data alone
packages/db Drizzle schema
packages/shared Zod schemas shared across the API boundary
pnpm install
pnpm gate # lint + typecheck + test — green before every commit
pnpm dev # all apps
pnpm lint:fix
Requires Node >= 24 and pnpm 9.
The agent's private key never leaves the client's environment, so AgentTrace cannot
forge or backdate a manifest. Storage holds the content but does not vouch for it —
tampering is caught by the on-chain root. The chain attests only that a byte string
existed in a given slot; signature validity is established off-chain by
packages/verify, which by design cannot reach the AgentTrace API.
MIT — see LICENSE.