TapeAPI 1.0 is officially live. Two days ago, when I first introduced TapeAPI, it was still just a pre-alpha: Let a Service on TapeOut prove “who said this.” But after building it out to 1.0, I realized it’s becoming something much more interesting: Every AI / Agent call can now come with a cryptographically signed receipt. For example, when you call an OpenAI- or Anthropic-compatible API: ▪ Who responded ▪ Which Request the response corresponds to ▪ Which bytes were returned ▪ The claimed token usage ▪ The price being charged All of that can be committed into a signed Receipt. The service gets its onchain identity from a TapeOut Circuit Container, while its Endpoint and Price Table can be published onchain. And for developers, almost nothing changes: Keep using the official OpenAI / Anthropic SDK. Just change the Base URL. For 1.0, there are a few updates I think matter most: 1. AI Signing Sidecar OpenAI Chat / Responses, Anthropic Messages, Embeddings, and Streaming are all supported. Claude Code and Codex have also been tested end-to-end. 2. new-api can plug directly into TapeAPI No modification to new-api itself. One docker compose up, and an existing AI relay / API provider can start attaching signed receipts to responses. In other words: Existing AI infrastructure doesn’t need to be rebuilt. 3. tapeapi-verify Claude Code / Codex users can verify every AI response locally. Under strict mode: Bad signature, mismatched Request, invalid Receipt → the response never enters the Agent. 4. MCP TapeAPI now has a public MCP endpoint: api.tapeapi.fun/mcp ▪ 8 tools, all returning signed Receipts ▪ TapeAPI is now listed in the Official MCP Registry If you already run your own MCP Server, you can also put the TapeAPI Signing Proxy in front of it. So instead of just: AI → MCP → Tool we can now have: AI → MCP → Signed Tool → Verifiable Receipt 5. Private Channels / Groups TapeOut Containers can now establish end-to-end encrypted channels and groups. The Relay only sees ciphertext. So TapeAPI is starting to cover the full machine-to-machine stack: ▪ Service Identity ▪ API ▪ MCP ▪ AI Receipts ▪ Machine Communication TAP-20 / 21 / 23 / 26 / 27 have also moved to Stable v1. Starting from 1.0, TapeAPI follows Semantic Versioning: Anything built against the 1.0 docs will remain compatible throughout 1.x. A true breaking change only happens in 2.0. One important boundary: A TapeAPI Receipt proves: “Who said what, in response to which request.” It does not prove that the provider actually ran the model it claims to have run. And it does not prove that the answer itself is true. What TapeAPI gives you is: Provenance + Accountability. And I increasingly think this becomes fundamental in an Agent Economy. If billions of Agents are going to call APIs, MCP servers, tools, and other Agents autonomously, machines cannot just exchange outputs. They also need to exchange: cryptographic proof of who is accountable for that output. That is how I currently think about TapeAPI. TapeAPI 1.0: github.com/BruceLanLan/tapea… GitHub: github.com/BruceLanLan/tapea… Website: tapeapi.fun And once again, credit to @Theairresearch for originating the TapeAPI idea. 10% of any future TapeAPI revenue will permanently go to the idea originator. No token. Just open source. #TapeOut Builder Culture 🦾

谢谢，感觉可以做的更逼真一点哈

同时拥有自己的手机号码和邮箱的 Agent 我是第一次见！🤯

作者更新简介和原理：