SUNGLASSES
Input firewall for AI agents: scan text and files before they're read. Local, no network by default.
作者提供的说明(英文原文)
Live honeypot threat intel: attacker IPs, IOCs, credentials, payloads, actors.
这三项永远不会合并成一个分数:意见、测量和人气是不同的东西。
我们还没有为它写通俗的介绍。下面是我们从注册中心读取到的事实。
选择你使用的 AI 应用,然后复制设置。
BLACKDOME_API_KEY, BLACKDOME_BASE_URL, BLACKDOME_TIMEOUT标为机密的值是你账户的钥匙。千万不要分享给别人,也不要粘贴到聊天里。
你需要
你用的是哪个应用?
在上方选择你的应用。我们会记住它,下次按钮会显示“安装到 …”。
打开 Claude Desktop → Settings → Developer → Edit Config,把下面的内容粘贴到 claude_desktop_config.json
{
"mcpServers": {
"blackdome": {
"command": "uvx",
"args": [
"blackdome-mcp"
],
"env": {
"BLACKDOME_API_KEY": "your-value",
"BLACKDOME_BASE_URL": "your-value",
"BLACKDOME_TIMEOUT": "your-value"
}
}
}
}在终端中运行这条命令
claude mcp add blackdome --env BLACKDOME_API_KEY=your-value --env BLACKDOME_BASE_URL=your-value --env BLACKDOME_TIMEOUT=your-value -- uvx blackdome-mcp运行前请先填写空着的值。
粘贴到项目中的 .cursor/mcp.json,或 ~/.cursor/mcp.json
{
"mcpServers": {
"blackdome": {
"command": "uvx",
"args": [
"blackdome-mcp"
],
"env": {
"BLACKDOME_API_KEY": "your-value",
"BLACKDOME_BASE_URL": "your-value",
"BLACKDOME_TIMEOUT": "your-value"
}
}
}
}浏览器会先询问是否打开应用,然后应用会显示设置并请你确认。密钥的值会留空,由你自己填写。
粘贴到项目中的 .vscode/mcp.json
{
"servers": {
"blackdome": {
"type": "stdio",
"command": "uvx",
"args": [
"blackdome-mcp"
],
"env": {
"BLACKDOME_API_KEY": "your-value",
"BLACKDOME_BASE_URL": "your-value",
"BLACKDOME_TIMEOUT": "your-value"
}
}
}
}浏览器会先询问是否打开应用,然后应用会显示设置并请你确认。密钥的值会留空,由你自己填写。
这个服务器在你的电脑上运行,而 ChatGPT 只能连接互联网上的服务器。请选择其他应用。
你的选择通过 Cookie 保存在这台设备上,无需账户。
BLACKDOME_BASE_URL envBLACKDOME_TIMEOUT env这些设置是根据作者在注册中心的条目生成的。如果 README 的写法不同,请以 README 为准。
它在你的电脑上运行;安装后,工具列表会显示在你的 AI 应用中。
我们检查的是任何人都能看到的信息;我们没有阅读代码。不适用的检查项不计入总数。 · 检查于 2026年11月9日
注册名称已通过 GitHub 账户 blackdome-ai 验证,该账户拥有此仓库
我们还没有从 GitHub 读取这个仓库
我们还没有从 GitHub 读取这个仓库
它在你的电脑上运行,没有互联网端点
在你的电脑上运行,改用本地设置
它在你的电脑上运行;我们只能读取互联网端点的工具列表
与注册中心里这个分类中所有活跃的服务器比较。只有当发布者拥有该仓库时,星标数才计入。
它在 1 项适用的基础检查中通过了 1 项。我们没有阅读代码。在添加密钥或令牌之前,请先看看这个服务器要求哪些权限,并使用只具备所需权限的密钥。
你需要填写:BLACKDOME_API_KEY
它在你的电脑上运行。你的 AI 应用会在需要时启动它,数据不会经过作者的服务器(除非它自己调用在线服务)。
GitHub 仓库最后更新于–,最新版本(0.2.0)于5个月前进入注册中心。
支持 MCP 的应用,例如 Claude Desktop、Claude Code、Cursor 和 VS Code。你的电脑上还需要有 Python (uv)。
同一分类中有:SUNGLASSES, KillerSkills, Draugr
同一分类中其他作者的服务器,按星标数排序。
如何看卡片:↓ 每周下载量i★ GitHub 星标i三大支柱:评价 · 检查 · 使用情况i更新 = 最新版本发布到注册中心的时间i
Input firewall for AI agents: scan text and files before they're read. Local, no network by default.
Security verdicts for agent skills and MCP servers, from sandbox runs, before you install them.
Security scanning for AI agents: SAST, SCA, secrets, IaC, DAST, ranked by real risk.
按原文从作者的仓库复制。PopNx MCP 没有审查或背书其内容。HTML、图片和脚本已被移除。链接会离开本站。
Canonical MCP server name: io.github.blackdome-ai/blackdome-mcp
Package version: 0.4.0
Give your AI agents direct access to read-only observed threat intelligence. Look up attacker IPs, browse indicators of compromise (IOCs), inspect captured credentials and malware payloads, compare samples deterministically, inspect bounded evidence clusters, and retrieve authenticated ThreatDrop case and workflow status — all from Claude, Cursor, or any MCP-compatible client.
Community tools are free and need no API key. The tool table identifies the product features required for paid intelligence.
One hosted streamable-HTTP endpoint for compatible remote MCP clients (Claude Desktop, the claude.ai web app, mobile, Cursor):
https://api.blackdome.ai/mcp
Free tools work with no key. To unlock the paid tiers (credential intelligence, payloads, actors, warboard, STIX export), get an API key at [https://blackdome.ai/pricing](https://blackdome.ai/pricing) and add it as an Authorization header:
{
"mcpServers": { "blackdome-cloud": { "url": "https://api.blackdome.ai/mcp", "headers": { "Authorization": "Bearer bd_your_key_here" } } } }
Use uvx (part of uv) to fetch and run the server on demand:
uvx blackdome-mcp
For a fixed install, use pip install blackdome-mcp.
The free community tools use no API key. To unlock paid tiers, get a key at [https://blackdome.ai/pricing](https://blackdome.ai/pricing).
Merge this into claude_desktop_config.json — ~/Library/Application Support/Claude/ on macOS, %APPDATA%\Claude\ on Windows — then restart Claude:
{
"mcpServers": { "blackdome": { "command": "uvx", "args": ["blackdome-mcp"], "env": { "BLACKDOME_API_KEY": "bd_your_key_here" } } } }
The
envblock is optional — omitBLACKDOME_API_KEYto run free public tools only.
One command — the key is stored in the MCP config, so there are no shell exports to maintain (a plain export BLACKDOME_API_KEY=... only lasts for that terminal session, and your paid tools would stop working in the next one):
claude mcp add blackdome -e BLACKDOME_API_KEY=bd_your_key_here -- uvx blackdome-mcp
For free tools only:
claude mcp add blackdome -- uvx blackdome-mcp
Add to ~/.cursor/mcp.json (global) or .cursor/mcp.json (per project):
{
"blackdome": { "command": "uvx", "args": ["blackdome-mcp"], "env": { "BLACKDOME_API_KEY": "bd_your_key_here" } } }
401.whoami) fail loudly with 401 Invalid API key — there is no silent fallback to free-tier results. Free tools keep working regardless of key validity.403 API key has expired / 403 Tenant account is inactive — again explicit errors, not degraded data.403; check your granted features any time with whoami.If a paid tool returns less than expected, run whoami first — it reports your plan, features, and live quota.
Free tools work with no key. Paid tools require an API key whose plan includes the listed feature.
| Tool | Tier | Description |
|---|---|---|
lookup_attacker_ip | Free | Full dossier for one attacker IP — events, protocols, credentials (passwords masked), MITRE, edge nodes |
top_attackers | Free | Most active attacker IPs over a window — pick one to drill into |
attack_map | Free | Recent geolocated attack events for a live map (limit ≥ 10) |
attack_heatmap | Free | Country-aggregated attack heatmap with centroids (limit ≥ 5) |
credential_preview | Free | Sample of recent credentials (masked server-side) + teaser totals |
verify_sigil | Free | Verify a BlackDome Sigil / audit record by id |
recent_iocs | Free | Browse recent redacted IOCs — type/severity filters (72h community delay, 25-row cap) |
ioc_trends | Free | Aggregated IOC trends — totals, breakdowns, daily new, top MITRE |
export_iocs | Free (json/csv) · Pro (stix) | Export the IOC feed; STIX bundle needs the stix_export feature |
search_credentials | Credential Intelligence or candidate Enterprise/OEM pilot (credential_intel or realtime_credential_feed) | Search the global credential corpus with PLAINTEXT passwords |
credential_stats | Credential Intelligence or candidate Enterprise/OEM pilot (credential_intel or realtime_credential_feed) | Aggregate credential stats — top usernames/passwords, breakdowns |
list_payloads | Requires api_access | List captured malware payloads, or fetch one by sha256 (VT/MB intel) |
get_actor | Requires api_access | List clustered threat actors, or fetch one actor's sessions |
warboard | Requires api_access | Sigil leaderboard with intrusion narratives + attacker command tails |
list_notable_sessions | Candidate Enterprise/OEM pilot (session_intel) | Ranked hand-keyed attacker sessions surfaced out of botnet noise |
get_session_transcript | Candidate Enterprise/OEM pilot (session_intel) | Structured command/output transcript for one attacker session |
list_detonations | Pro (detonation_intel) | Malware detonation list with verdicts, Magika labels and IOC counts |
get_detonation_report | Pro (detonation_intel) | Full detonation report with behavior, IOCs, artifact classification and report availability |
get_artifact | Pro (detonation_intel) | Artifact dossier with linked detonation, IOCs and session identifiers only |
explain_ioc | Analyst (ioc_explanation) | Observed evidence for an IP, domain, or URL |
compare_samples | Pro (sample_compare) | Deterministic comparison of two SHA-256 samples |
get_evidence_cluster | Pro (genome_clusters) | Bounded evidence cluster for a SHA-256, IP, domain, or URL |
get_reproducible_novelty | Pro (reproducible_novelty) | Reproducible behavior assessment for a SHA-256, IP, domain, or URL |
get_threatdrop_case | ThreatDrop Pro or candidate Enterprise pilot (api_access) | Authenticated account's sanitized case state; separate ThreatDrop product entitlement |
get_takedown_status | ThreatDrop Pro or candidate Enterprise pilot (api_access) | Authenticated account's public case counts and takedown workflow status, including replies, outcome, and timeline |
whoami | Any key | Check your tenant, plan, features and live quota |
Feature gates: tool availability is enforced by the returned plan features, including api_access, stix_export, credential_intel or realtime_credential_feed, detonation_intel, session_intel, ioc_explanation, sample_compare, genome_clusters, and reproducible_novelty. ThreatDrop tools require a separate ThreatDrop product entitlement. See pricing.
Once connected, try asking your AI assistant:
whoami)The package runs locally over stdio; the separately hosted streamable-HTTP endpoint takes the key as an Authorization: Bearer header instead. The registry manifest declares only the package's stdio transport.
| Variable | Required | Default | Description |
|---|---|---|---|
BLACKDOME_API_KEY | No | — | Bearer API key. Free tools work without it; paid tools require it |
BLACKDOME_BASE_URL | No | https://api.blackdome.ai | API base URL |
BLACKDOME_TIMEOUT | No | 15 | Request timeout in seconds |
The free community tier is capped at roughly 30 requests/minute and 100 requests/day, and community IOC data carries a 72-hour freshness delay. Current paid-plan limits are returned by whoami; candidate Enterprise/OEM pilots are contact-led and do not have fixed public limits. When you hit a limit the server returns a clear 429 error with retry timing.
"command": "uvx", "args": ["blackdome-mcp"] as shown above. For a pip installation, run which blackdome-mcp (macOS/Linux) or where blackdome-mcp (Windows) and use the full path as the command.whoami tells you exactly what your key grants.lookup_attacker_ip tool masks captured passwords to ******** before returning them; credential_preview is masked server-side. Plaintext passwords are returned only by the paid search_credentials tool, which requires credential_intel or realtime_credential_feed.MIT