
BountyRadar MCP
io.github.PanStoriesPublished by GitHub account PanStoriesv1.0.1Updated Oct 27, 2026
Ranked feed of AI-agent-solvable bounties, scored for solvability and competition.
Overview
BountyRadar MCP serves a ranked, scored feed of AI-agent-solvable open-source and security bounties so an assistant can pick targets worth attempting.
- What you can do
- Fetch a ranked, filtered bounty feed by category, minimum reward, maximum competition, source, freshness and solvability
- Search bounties by keyword across titles, descriptions, tags and repositories
- Read a single bounty's full record and explainable score breakdown, or score one bounty directly
- List the source registry with status, and get counts by source, category and reward band
- Save a filter profile with a poll descriptor for repeat polling
- Read feed, stats and source resources, and use a daily bounty brief prompt
- When to use it
- Useful when an agent works the open-source bounty economy and needs triage rather than discovery: ranking candidates by solvability, freshness and competition, and filtering out honeypot listings before spending compute.
- Requirements
- The hosted endpoint is a remote Streamable HTTP MCP server requiring an Apify token sent as an Authorization Bearer header; OAuth is declared. Pay-per-event billing applies to most tools. Alternatively, self-host the MIT-licensed Node 22 stdio server from source; local use is free.
Installation
In SourceWeft
- Open BountyRadar MCP in the dashboard and add it to a workspace.
- Connect the required credentials. They are stored privately and never shown on public pages.
- Enable the server for the chats that should use its tools.
Web executable via Streamable HTTP. Remote servers run from the web runtime once configured in a workspace.
Use with an agent
Paste this prompt into Claude Code, Codex, Cursor or another agent — or into SourceWeft chat. The agent reads this server's install guide, shows you its publisher, endpoint and sign-in, and adds it to your client once you agree. Keys never go into a command.
Read https://sourceweft.com/mcp/io.github.PanStories%2Fbountyradar-mcp/install.md and connect the MCP server it describes to my client. Before changing any config, show me its publisher, endpoint and sign-in method and wait for my OK. Never put an API key in a command or in a file I would commit; ask me where it should come from.Other MCP clients
Add this to your client's mcpServers config. The client completes OAuth sign-in on first connect.
{
"mcpServers": {
"bountyradar": {
"type": "http",
"url": "https://neeenja--bountyradar-mcp.apify.actor/mcp"
}
}
}README
BountyRadar MCP
The judgment layer for the agent-bounty economy. BountyRadar aggregates AI-agent-solvable bounties across Opire, BountyHub, ClawHunt, Algora, GitHub, and validated security programs — then scores each for agent-solvability, freshness, and competition and serves a ranked, agent-consumable feed over MCP.
Positioning: an intelligence / alerting layer — NOT a marketplace. No fund custody, no KYC, no cross-border settlement. Public OSS bounties are agent-saturated; the scarce skill is target selection, not discovery. BountyRadar is the selection layer.
Languages: English · 简体中文 · 繁體中文
English
Current version: 1.0.1
Why this exists
- Public open-source bounties are agent-saturated — a single Algora bounty has drawn 8–158 competing agent attempts within hours.
- The bottleneck is triage / judgment, not discovery — so the moat is the scoring layer, not scraping.
- BountyRadar ranks and de-noises so an agent spends compute only on targets it can actually win.
Quick start (hosted MCP endpoint — recommended)
The hosted endpoint is a Streamable HTTP MCP server with Pay-Per-Event billing (no monthly fee). Connect it with any MCP client:
Claude Desktop / Claude Code / WorkBuddy / any MCP client (remote HTTP):
initialize and tools/list are free; only value events are charged (see Pricing).
No install step and no npm required — the endpoint is the product. One-liner for the common clients:
Run locally (stdio, MIT — clone and self-host)
The full source (Node 22, zero-dependency stdio server) is MIT-licensed and lives on GitHub: https://github.com/PanStories/bountyradar-mcp. Clone it and run:
MCP client config (local stdio):
Tools
Resources: bountyradar://feed/latest, bountyradar://stats, bountyradar://sources
Prompt: daily_bounty_brief — briefs an agent on today's best targets.
Scoring (the moat — explainable, deterministic)
A free LLM (Cheapest-LLM Router / Cloudflare Workers AI) only assists task_type / agent_solvable at refresh time; scoring itself is deterministic and needs no training.
Honeypot guard
Bounties flagged honeypot: true (e.g. "reserved for interview loop") are excluded from the default feed so agents don't waste compute or burn reputation. Toggle with include_honeypot: true.
Monetization — Pay-Per-Event (no monthly fee)
initialize / tools/list / list_sources / get_bounty_detail are free. Charged events:
Local / self-hosted use is free forever.
Price update (floor correction): all charged events raised to $0.005 effective 2026-11-06; the current $0.0005 / $0.0010 prices are grandfathered until then. This is a floor correction (the previous $0.0005 was 10× below the $0.005 PPE floor), not a margin play.
Self-host / deploy (Apify, PPE)
Standby mode (120s / 512MB) keeps idle cost near zero. The actor's main process binds the HTTP port itself (Standby-safe) and answers the container readiness probe.
Dev & verification
Optional add-ons (same catalog)
- Website + subscriber alerts —
node website/build.mjs→ a zero-dependency static site + Cloudflare Pages Functions (KV subscribe, guarded CSV export);scripts/digest.mjsrenders per-subscriber email/Slack digests. - Live refresh —
scraper/is a standalone Apify actor (orscripts/refresh.mjslocally) that collects real bounty issues and writes todata/live/, never polluting the curateddata/bounties/. SetBR_INCLUDE_LIVE=1to merge at runtime.
Honesty: only issues that are actually bounties (a stated
$reward or abountylabel) are kept; unfunded "good first issues" are noise and excluded. Opire / Algora / BountyHub require platform keys; ClawHunt is apartner_candidateand is intentionally not scraped (index upstream, don't compete).
License
MIT — see the LICENSE file. Free to use the hosted endpoint per the pricing table above.
Privacy
Read-only and stateless — no accounts, no personal data collected. An optional webhook_url is held in memory only; optional classification may send public bounty text to a configured LLM endpoint. See PRIVACY.md.
简体中文
Current version: 1.0.1
Agent 赏金经济的"判断层"。 BountyRadar 跨 Opire / BountyHub / ClawHunt / Algora / GitHub 及已核验安全计划聚合"AI agent 可解"的赏金,对每条做 agent 可解性 / 新鲜度 / 竞争度 评分,并通过 MCP 输出一份排好序、agent 可直接消费的 feed。
定位: 情报 / 告警层——不是市场 / 结算层。不碰资金托管、KYC、跨境结算。公开 OSS 赏金已被 agent 淹没;稀缺能力是"选目标",不是"发现目标"。BountyRadar 就是这一层。
为什么做这个
- 公开开源赏金已被 agent 农场淹没——单条 Algora 赏金几小时内就涌进 8–158 个抢单 agent。
- 瓶颈在分流 / 判断,不在发现——所以护城河是评分层,不是爬虫。
- BountyRadar 排序 + 去噪,让 agent 只把算力花在真正能赢的目标上。
快速开始(托管 MCP 端点——推荐)
托管端点是**按事件付费(PPE)**的 Streamable HTTP MCP 服务器,无月费。任何 MCP 客户端均可接入:
Claude Desktop / Claude Code / WorkBuddy / 任意 MCP 客户端(远程 HTTP):
initialize 与 tools/list 免费;仅对价值事件计费(见「商业化」)。
无需安装、无需 npm —— 端点本身就是产品。常见客户端一行搞定:
本地运行(stdio,MIT 开源,可克隆自托管)
完整源码(Node 22、零依赖 stdio 服务器)以 MIT 许可证开源在 GitHub:https://github.com/PanStories/bountyradar-mcp。克隆后运行:
MCP 客户端配置(本地 stdio):
工具
资源: bountyradar://feed/latest、bountyradar://stats、bountyradar://sources
提示词: daily_bounty_brief(briefing 今日最佳目标)
评分(护城河——可解释、确定性)
免费 LLM(Cheapest-LLM Router / Cloudflare Workers AI)仅在 refresh 时辅助判定 task_type / agent_solvable;评分本身是确定性、无需训练。
蜜罐守卫
标记为 honeypot: true 的赏金(如"面试专用")默认移出 feed,避免 agent 浪费算力或毁声誉。include_honeypot: true 可切回显示。
商业化——按事件付费(无月费)
initialize / tools/list / list_sources / get_bounty_detail 免费。计费事件:
本地 / 自托管永久免费。
价格更新(地板修正): 所有计费事件于 2026-11-06 起统一提至 $0.005;现行 $0.0005 / $0.0010 价格祖父保留至该日。这是地板修正(此前 $0.0005 低于 $0.005 地板 10 倍),非涨价牟利。
自托管 / 部署(Apify,PPE)
Standby 模式(120s / 512MB)让闲置成本接近零。Actor 主进程自行绑定 HTTP 端口(Standby 安全)并响应容器就绪探针。
开发与验证
可选增强(同一份目录数据)
- 网站 + 订阅告警——
node website/build.mjs生成零依赖静态站 + Cloudflare Pages Functions(KV 订阅、受保护的 CSV 导出);scripts/digest.mjs渲染按订阅者定制的邮件 / Slack 摘要。 - 实时刷新——
scraper/是独立的 Apify actor(或本地scripts/refresh.mjs),采集真实赏金 issue 并写入data/live/,绝不污染 curated 的data/bounties/。设BR_INCLUDE_LIVE=1可在运行时合并。
诚实声明:只保留真正是赏金的内容(写明
$金额或带bounty标签);无赏金的 "good first issue" 是噪音,已排除。Opire / Algora / BountyHub 需平台密钥;ClawHunt 是partner_candidate,刻意不采集(上游索引,不竞争)。
许可证
MIT——详见 LICENSE。按上表计费使用托管端点。
繁體中文
Current version: 1.0.1
Agent 賞金經濟的「判斷層」。 BountyRadar 跨 Opire / BountyHub / ClawHunt / Algora / GitHub 及已核驗安全計畫聚合「AI agent 可解」的賞金,對每條做 agent 可解性 / 新鮮度 / 競爭度 評分,並透過 MCP 輸出一份排好序、agent 可直接消費的 feed。
定位: 情報 / 警報層——不是市場 / 結算層。不碰資金託管、KYC、跨境結算。公開 OSS 賞金已被 agent 淹沒;稀缺能力是「選目標」,不是「發現目標」。BountyRadar 就是這一層。
為什麼做這個
- 公開開源賞金已被 agent 農場淹沒——單條 Algora 賞金幾小時內就湧進 8–158 個搶單 agent。
- 瓶頸在分流 / 判斷,不在發現——所以護城河是評分層,不是爬蟲。
- BountyRadar 排序 + 去噪,讓 agent 只把算力花在真正能贏的目標上。
快速開始(託管 MCP 端點——推薦)
託管端點是**按事件付費(PPE)**的 Streamable HTTP MCP 伺服器,無月費。任何 MCP 用戶端均可接入:
Claude Desktop / Claude Code / WorkBuddy / 任意 MCP 用戶端(遠端 HTTP):
initialize 與 tools/list 免費;僅對價值事件計費(見「商業化」)。
無需安裝、無需 npm —— 端點本身就是產品。常見客戶端一行搞定:
本機執行(stdio,MIT 開源,可克隆自架)
完整原始碼(Node 22、零依賴 stdio 伺服器)以 MIT 許可證開源在 GitHub:https://github.com/PanStories/bountyradar-mcp。克隆後執行:
MCP 用戶端設定(本機 stdio):
工具
資源: bountyradar://feed/latest、bountyradar://stats、bountyradar://sources
提示詞: daily_bounty_brief(briefing 今日最佳目標)
評分(護城河——可解釋、確定性)
免費 LLM(Cheapest-LLM Router / Cloudflare Workers AI)僅在 refresh 時輔助判定 task_type / agent_solvable;評分本身是確定性、無需訓練。
蜜罐守衛
標記為 honeypot: true 的賞金(如「面試專用」)預設移出 feed,避免 agent 浪費算力或毀聲譽。include_honeypot: true 可切回顯示。
商業化——按事件付費(無月費)
initialize / tools/list / list_sources / get_bounty_detail 免費。計費事件:
本機 / 自架永久免費。
價格更新(地板修正): 所有計費事件於 2026-11-06 起統一提至 $0.005;現行 $0.0005 / $0.0010 價格祖父保留至該日。這是地板修正(此前 $0.0005 低於 $0.005 地板 10 倍),非漲價牟利。
自架 / 部署(Apify,PPE)
Standby 模式(120s / 512MB)讓閒置成本接近零。Actor 主程序自行繫結 HTTP 埠(Standby 安全)並回應容器就緒探針。
開發與驗證
選用增強(同一份目錄資料)
- 網站 + 訂閱警報——
node website/build.mjs產生零依賴靜態站 + Cloudflare Pages Functions(KV 訂閱、受保護的 CSV 匯出);scripts/digest.mjs渲染按訂閱者定製的郵件 / Slack 摘要。 - 即時更新——
scraper/是獨立的 Apify actor(或本機scripts/refresh.mjs),採集真實賞金 issue 並寫入data/live/,絕不污染 curated 的data/bounties/。設BR_INCLUDE_LIVE=1可在執行時合併。
誠實聲明:只保留真正是賞金的內容(寫明
$金額或帶bounty標籤);無賞金的 "good first issue" 是雜訊,已排除。Opire / Algora / BountyHub 需平台金鑰;ClawHunt 是partner_candidate,刻意不採集(上游索引,不競爭)。
授權
MIT——詳見 LICENSE。按上表計費使用託管端點。
Source: README.md at commit 8f7f56f
Tools
0Version history
1- v1.0.1LatestOct 9, 2026
Score
85% of the applicable checks met
How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.
- Reached in the last 30 daysRequired25 pts
- Auth declaredRequired10 pts
- Tools listed15 pts
Answer tools/list without a credential, or ship a package that starts in a sandbox without one.
- DescriptionRequired15 pts
- Logo10 pts
- Documentation15 pts
- License10 pts