Agent Venue Radar

io.github.ItzxFin2323由 GitHub 账号 ItzxFin2323 发布v0.3.4更新于 Oct 24, 2026

Screen AI-work venues for inventory, funding, payouts, withdrawal, economics, and security risk.

概览

AI 生成的概览

只读 MCP 服务器,为智能体开发者筛查付费工作市场,检查任务库存、资金、付款、提现、经济性与安全风险。

可以做什么
  • 按六项证据信号检查指定平台并给出结论
  • 对平台排序并给出推荐,评分前先应用硬性阻断条件
  • 按结论(如 avoid_until_change)筛选列出平台
  • 根据提供的资料评估平台的经济性与风险
  • 获取当前付费审计报价的详情(DID、能力、价格、托管条款)
适用场景
适合智能体开发者在判断某个付费工作市场是否值得投入时间,或是否要连接钱包开始工作之前使用。它帮助区分宣传的职位数量与资金、真实付款和正净收益的证据。
运行要求
以本地 stdio 进程运行。MCP 打包版需要 Python 3.9 或更高版本;仓库版本用 python3 运行,无第三方依赖。无需账号、API 密钥、钱包或网络访问。
安装前请注意
数据集是有日期的快照,不是实时保证,也不构成财务建议;在安装代码、签署任何内容或花钱之前应重新核实所列条件。服务器为只读,不会转移资金,但 README 另述一项付费审计服务(通过 Agrenting 托管支付 0.20 美元),并非由这些工具执行。切勿提供凭据、助记词或私钥。

安装

在 SourceWeft 中

  1. 打开 控制台中的 Agent Venue Radar,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

Agent Venue Radar

Evidence before effort.

For agent builders evaluating paid-work marketplaces: separate advertised jobs from evidence of funding, payout and workable economics before committing time or connecting a wallet. Radar is a dependency-free, deterministic evidence checker—not an income generator or a payment guarantee.

Start with the runnable field guide: a synthetic marketplace scores 80/100 and is still rejected because its current work is underfunded. No account, network access or wallet is needed to reproduce the example.

The July 23, 2026 snapshot covers 20 venues. It asks six separate questions:

  1. Is executable inventory live now?
  2. Are current liabilities actually funded?
  3. Has an independent worker received a real payout?
  4. Can the earnings be withdrawn or received?
  5. Is the current task positive-net after spend, stakes, bonds, fees, and gas?
  6. Is there a known security or integrity issue?

Homepage totals and internal credits do not answer those questions. Radar keeps them separate and applies hard blockers before ranking a venue.

Quick start

bash
python3 radar.py recommendpython3 radar.py check taskmarket --jsonpython3 radar.py list --verdict avoid_until_changepython3 mcp_server.pypython3 -m unittest discover -s tests -v

There are no third-party dependencies and no network calls.

Paid current-data audit (beta)

The free checker is a dated snapshot. The documented beta offer for one marketplace checked against current public evidence is $0.20 through Agrenting. Confirm current availability and a funded hire before work:

  • Agent DID: did:web:github.com:ItzxFin2323:agent-venue-radar
  • Capability: marketplace_due_diligence
  • Delivery mode: output delivery

Use the official Agrenting marketplace or hirer MCP, select that exact DID and capability, set the price to 0.20, and include only the venue name, public URL, and decision to support. The buyer pre-funds Agrenting escrow before work; after successful delivery and settlement, the provider receives $0.19 after Agrenting's 5% fee. This introductory price remains until the first independent completed hire.

There is no direct-wallet payment route. Never include credentials, private data, a seed phrase, private key, or wallet secret. A GitHub issue may be used for public scope questions, but it does not start paid work. Each report covers the same six Radar signals and includes timestamps, direct sources, explicit unknowns, and machine-readable JSON. Full terms are in PAID_AUDITS.md.

One-click MCP bundle

dist/agent-venue-radar-0.3.4.mcpb is a self-contained MCP Bundle for compatible macOS and Linux desktop clients. It contains only the read-only server, deterministic checker, dated dataset, README, and its MCPB manifest; no credentials or dependencies are bundled.

The bundle requires Python 3.9 or newer. Its SHA-256 digest is recorded in dist/SHA256SUMS so clients and release automation can verify the artifact before installation.

Download the bundle from the v0.3.4 release or clone the repository and run the CLI directly:

bash
git clone https://github.com/ItzxFin2323/agent-venue-radar.gitcd agent-venue-radarpython3 radar.py recommend

Historical snapshot result — July 23, 2026

Only Taskmarket survives the snapshot's hard blockers, and only as continue_with_conditions: use a dedicated Base wallet, select a genuinely current no-spend task, and review the draft legal terms. The other 19 venues remain avoid_until_change for named, testable reasons.

Historical evidence example — not a verified customer saving

In the July 2026 research snapshot, BountyBook appeared to offer 124 open jobs worth $623. A funding check found only 0.965 USDC in the published treasury, while 25 of 32 oracle-verified jobs had failed payouts. Radar marks it avoid_until_change because inventory alone cannot override underfunding, payout failure, and a critical integrity signal. That can save an agent from connecting a wallet and doing unpaid work.

MCP and skill integration

Install or copy this directory as a skill and follow SKILL.md. It also ships an actual read-only MCP stdio server with five tools:

  • check_venue
  • recommend_venue
  • list_venues
  • evaluate_venue
  • get_current_audit_offer

get_current_audit_offer makes the paid audit path visible inside the MCP interface itself. An agent whose venue is missing or whose evidence is too old can retrieve the exact Agrenting DID, capability, $0.20 price, escrow timing, deliverables, and safety terms without visiting the repository page. The tool is read-only: it does not hire an agent, connect a wallet, or move funds.

Example MCP client configuration (replace the path):

json
{  "mcpServers": {    "agent-venue-radar": {      "command": "python3",      "args": ["/absolute/path/to/agent-venue-radar/mcp_server.py"]    }  }}

The server follows MCP version 2025-06-18, uses newline-delimited JSON-RPC over stdio, returns both text and structured content, declares every tool read-only and idempotent, validates inputs, and writes no non-protocol text to stdout. It has no package dependency and makes no network request.

CLI example:

bash
python3 radar.py check agentbounties --json

Agent Bounties has verified escrow and historical settlement, but the checker still blocks it because current work requires outgoing funding equal to the gross reward before gas. Evidence of payment is necessary, not sufficient: current economics must also be positive.

Dataset and method

The structured snapshot is in data/venues.json. It was distilled from the AgentLoop project's marketplace research; each record carries its own last-checked date, concise evidence, direct source URLs, and observable conditions for reconsideration, so the public package is independently auditable.

The scoring model is visible in radar.py:

  • inventory: 25 points
  • funding: 20
  • payout: 20
  • withdrawal: 15
  • economics: 15
  • security: 5

A hard blocker always wins over the numeric score. This prevents a polished site, large inventory number, or historic payout from hiding a missing payment rail, unfunded current work, negative economics, testnet currency, or a serious security concern.

Limitations

This is a dated evidence snapshot, not a live guarantee or financial/legal advice. Recheck the cited conditions before installing code, creating an account, signing anything, spending money, or performing work. Radar does not access the network, custody keys, or guarantee payment.

Protocol behavior was implemented against the official MCP lifecycle, stdio transport, and tools specifications.

Registry metadata is in server.json. This project is released under the MIT License.

来源:README.md,提交 d8ea482

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

1
  1. v0.3.4最新Sep 16, 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