Fairseal

io.fairseal由 fairseal.io 发布v0.2.4更新于 Nov 4, 2026

Verifiable entropy, provably-fair games, and receipt verification for AI agents (FairSeal API).

概览

AI 生成的概览

FairSeal MCP 服务器通过 FairSeal API 为助手提供可验证熵、可证明公平的抽取凭证以及凭证校验能力。

可以做什么
  • 从公开的 drand 信标轮次获取可验证随机数与熵
  • 提交随机抽取的承诺,并按已承诺规则解析结果
  • 创建、公证并锚定结果凭证,证明结果在揭晓前已固定
  • 离线校验凭证并查询使用统计
  • 向外部链提交锚定,以获得可独立核查的时间证明
适用场景
当助手需要执行或审计他人事后可核查的随机分配时使用,例如抽卡、掉落、匹配、赛事种子、铸造顺序,或任何需要证明结果在揭晓前已承诺的场景。
运行要求
以 stdio 在本地运行,通常通过 npm 包 @fairseal/mcp-server 安装。受限工具需要 FAIRSEAL_API_KEY 中的 API 密钥;可选用 FAIRSEAL_ANCHOR_API_KEY 和 FAIRSEAL_NOTARY_API_KEY 覆盖,未设置时回退到前者。FAIRSEAL_API_URL 可指向其他 API 地址。需要访问 FairSeal API 与 drand 信标的网络。
安装前请注意
API 密钥属于机密,可访问受限工具;没有密钥时这些工具返回 401。锚定与公证会向外部服务和公开链写入记录,README 还提到通过 x402 结算小额支付,调用可能产生费用。凭证与承诺会提交给第三方并可被公开核查。

安装

在 SourceWeft 中

  1. 打开 控制台中的 Fairseal,将其添加到工作区。
  2. 为需要使用其工具的对话启用该服务。

Desktop only,通过 STDIO。 STDIO 服务会启动本地进程,因此需要 SourceWeft 桌面宿主。

其他 MCP 客户端

参照 仓库 中的启动说明。

README

Fairseal 🦭

Provably fair selections for games, loot boxes, and any randomized allocation.

Fairseal creates verifiable receipts (CSRs) proving that a selection was committed before the outcome was knowable. Every loot drop, gacha pull, and randomized selection — independently verifiable, trustlessly.

[@fairseal/commit] [@fairseal/core] [graded]


Try It Now

bash
npx @fairseal/commit

One command. A verifiable gacha pull in your terminal. No setup.

Quick Start

bash
npm install @fairseal/commit
typescript
import { createCommitment, resolveCommitment, createReceipt, verifyReceipt } from '@fairseal/commit';
// 1. Lock your loot table BEFORE the outcome is knowableconst commitment = createCommitment({  rule: JSON.stringify(LOOT_TABLE),  inputs: [playerId, pullNumber.toString()],  revealAfter: 10,});
// 2. Resolve with verifiable entropy from drandconst resolution = await resolveCommitment(commitment);
// 3. Your algorithm + verifiable entropy = provably fair dropconst drop = yourWeightedPull(LOOT_TABLE, resolution.beaconRandomness);
// 4. Package as a receipt anyone can verifyconst receipt = createReceipt(commitment, resolution);const proof = await verifyReceipt(receipt); // ✅ PARTIAL

How It Works

  1. Commit — Lock your drop table + player ID into a hash before the randomness exists
  2. Entropy — Wait for a public drand beacon round (Cloudflare + Protocol Labs)
  3. Resolve — Derive the selection from beacon entropy + your committed rule
  4. Verify — Anyone can re-run every step. drand beacon signatures are BLS12-381-verified; FairSeal receipt signatures use EIP-191 (secp256k1) / Ed25519 — receipts are never BLS-signed. No trust required.

Performance

OperationTime
Commitment creation0.006ms (170,000+/sec)
Beacon fetch~200ms
drand beacon verification (BLS12-381)~150ms
Receipt size0.9 KB

Client-side. No server. No account. No fees for PARTIAL verification.

Packages

PackageDescription
@fairseal/gameGame API for RGS studios. Slots, gacha, loot boxes. (v0.1 ALPHA)
@fairseal/commitCommitted selection receipts. Core library.
@fairseal/coreCryptographic primitives, signing, Merkle trees.
@fairseal/verifyIndependent receipt verification.
@fairseal/mcp-serverMCP server for AI agents. Verifiable entropy, provably-fair games, cryptographic receipt verification via Model Context Protocol.

Use Cases

ScenarioHow
Gacha / loot boxEvery pull has a receipt. Players verify drops match published rates.
Boss loot dropsDrop table committed per encounter. No ninja-nerfing.
MatchmakingPairing algorithm committed. Players verify no favoritism.
Tournament bracketsSeeding provably random, not rigged.
NFT mint orderQueue provably fair — no insider front-running.
Slot machinesSeed Commitment Model — verifiable spins, no nonce chain.

Result Notarization vs. Activity Receipts

Most "agent receipts" today are activity receipts: the platform logs what an agent did, and you trust whoever wrote the log. Fairseal receipts are result notarization — a different trust model:

Activity receipts (platform logs)Fairseal result notarization
Who do you trust?The operator writing the logNo one — receipts verify offline
What is proven?That an action was recordedThat a result was committed before the outcome was knowable
Time proofServer timestamp (operator-controlled)External anchor on Base (independently checkable)
Tamper evidenceDepends on log integrityHash-anchored; any modification breaks verification

Related standards: payment-authorization protocols like Google/Mastercard AP2 (mandates), Visa TAP, and OpenAI/Stripe ACP prove that an agent may act and who it is. Fairseal proves what happened — that a result was committed and hasn't been altered since. Fairseal settles micropayments over Coinbase's x402 and does not implement AP2/TAP/ACP; the layers are complementary.

Links

Listings

License

MIT — Fairseal

来源:README.md,提交 8ff9498

工具

0
SourceWeft 尚未在沙箱中启动过该软件包,因此其工具未知。

版本历史

2
  1. v0.2.4最新Sep 30, 2026
  2. v0.2.3Sep 27, 2026

评分

已满足适用检查项的 50%

该评分衡量此收录条目的完整程度,依据注册中心条目、端点与公开来源计算得出。它不是对服务器本身的评价,也不表示服务器安全或当前可用:可连通性是带日期的检查结果,评分也不包含安装量或用户评价。

  • 近 30 天内可连通25 分

    请保持端点能够响应 MCP 握手;SourceWeft 每 30 天探测一次。

  • 已声明认证方式10 分

    请在注册中心条目中声明凭据请求头,或保持端点能够响应探测,以便判定其认证方式。

  • 已列出工具15 分

    请让服务器在无凭据时响应 tools/list,或发布可在沙箱中无凭据启动的软件包。

  • 描述必需15 分
  • 图标10 分
  • 文档15 分
  • 许可证10 分