AgentDB
Real-time curated knowledge API for AI agents. Updated Mon/Wed/Fri from 31 sources covering AI/tech, startups, alternative markets, and emerging markets — no scraping or storage required.
README
AgentDB
Real-time curated knowledge API for AI agents.
AgentDB is a knowledge base updated Mon/Wed/Fri with summaries from 31 curated sources spanning AI/tech, startups, alternative markets, and emerging markets (Africa & Asia). Connect it to your AI agent so it always has fresh context — without you having to manage scraping, summarisation, or storage.
Quick start
1. Get an API key
curl -s -X POST https://agentdb-production-9ba0.up.railway.app/v1/auth/register \
-H "Content-Type: application/json" \
-d '{"email": "you@example.com", "name": "Your Agent"}'
Response:
{
"api_key": "adb_xxxxxxxxxxxxxxxxxxxx",
"tier": "trial",
"trial_expires_at": "2026-04-21T07:00:00",
"message": "Welcome to AgentDB. Your 3-day trial has started."
}
Store your key — it's shown once.
2. Fetch the latest knowledge
curl https://agentdb-production-9ba0.up.railway.app/v1/knowledge/latest \
-H "X-API-Key: adb_xxxxxxxxxxxxxxxxxxxx"
MCP server (recommended for Claude)
The AgentDB MCP server exposes two tools directly inside Claude Code or Claude Desktop: get_latest_knowledge and search_knowledge.
Install with uv (recommended)
uv handles Python version management automatically — no need to install Python 3.10+ manually.
curl -LsSf https://astral.sh/uv/install.sh | sh
Claude Code — global config
Add to ~/.claude.json (works in every Claude Code session):
{
"mcpServers": {
"agentdb": {
"type": "stdio",
"command": "/path/to/uv",
"args": ["run", "/path/to/agentdb/mcp/server.py"],
"env": {
"AGENTDB_API_KEY": "adb_xxxxxxxxxxxxxxxxxxxx"
}
}
}
}
Find your uv path with which uv. Find the server path with realpath mcp/server.py from this repo.
Claude Code — project config
Add a .mcp.json to your project root:
{
"mcpServers": {
"agentdb": {
"type": "stdio",
"command": "uv",
"args": ["run", "/absolute/path/to/agentdb/mcp/server.py"],
"env": {
"AGENTDB_API_KEY": "adb_xxxxxxxxxxxxxxxxxxxx"
}
}
}
}
Claude Desktop
Add to ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"agentdb": {
"command": "uv",
"args": ["run", "/absolute/path/to/agentdb/mcp/server.py"],
"env": {
"AGENTDB_API_KEY": "adb_xxxxxxxxxxxxxxxxxxxx"
}
}
}
}
Run the MCP server manually (test it)
AGENTDB_API_KEY=adb_xxxxxxxxxxxxxxxxxxxx uv run mcp/server.py
Available MCP tools
| Tool | Tier | Description |
|---|---|---|
get_latest_knowledge |
Trial + Pro | Fetch the N most recent items, optionally filtered by tags or content type |
search_knowledge |
Pro | Semantic vector search — find items most relevant to a natural language query |
REST API reference
Base URL: https://agentdb-production-9ba0.up.railway.app
All endpoints (except /health, /v1/auth/register, and /v1/knowledge/sources) require:
X-API-Key: adb_xxxxxxxxxxxxxxxxxxxx
Knowledge
| Method | Path | Tier | Description |
|---|---|---|---|
GET |
/v1/knowledge/latest |
Trial + Pro | Latest items; supports limit, page, tags, content_type |
GET |
/v1/knowledge/search?q=... |
Pro | Semantic search |
GET |
/v1/knowledge/{id} |
Trial + Pro | Single item by ID |
GET |
/v1/knowledge/sources |
Public | List of all scraped sources |
Query parameters — /latest
| Param | Type | Default | Description |
|---|---|---|---|
limit |
int | 20 | Items per page (1–100) |
page |
int | 1 | Page number |
tags |
string | — | Comma-separated tag filter, e.g. ai,markets |
content_type |
string | — | article, video, research, or data |
Example response item
{
"id": "3e7c224c-...",
"title": "Quantum Jamming and the Search for Principles Deeper Than Quantum Mechanics",
"content_type": "article",
"summary": "Researchers are exploring whether cryptographic protocols...",
"body": {
"category": "science_research",
"key_points": ["...", "..."],
"source_name": "Quanta Magazine"
},
"tags": ["quantum-mechanics", "cryptography", "causality"],
"confidence": 0.92,
"relevance_score": 0.92,
"published_at": "2026-04-18T11:40:55+00:00"
}
Auth
| Method | Path | Description |
|---|---|---|
POST |
/v1/auth/register |
Register and get a trial API key |
GET |
/v1/auth/me |
Inspect your key (tier, usage, expiry) |
Sources
AgentDB ingests from 31 sources, updated Mon/Wed/Fri at 07:00 UTC.
Focus: AI/tech, startups/IPO, alternative markets, and emerging markets (Africa & Asia). No legacy wire services.
Podcasts (8)
| Source | Category |
|---|---|
| Lex Fridman Podcast | technology_ai |
| Dwarkesh Podcast | technology_ai |
| Hard Fork | technology_ai |
| This Week in Tech | technology_ai |
| Acquired | startups_ipo |
| How I Built This | startups_ipo |
| All-In Podcast | market_news_alternative |
| Prof G Markets Podcast | market_news_alternative |
Blogs (22)
| Source | Category |
|---|---|
| MIT Technology Review | technology_ai |
| Ars Technica | technology_ai |
| IEEE Spectrum | technology_ai |
| Hacker News | startups_ipo |
| Y Combinator Blog | startups_ipo |
| The Verge | startups_ipo |
| Entrepreneur | startups_ipo |
| TechCrunch | startups_ipo |
| The Hindu Business Line | emerging_markets_asia |
| Mint (India) | emerging_markets_asia |
| Zero Hedge | market_news_alternative |
| Wolf Street | market_news_alternative |
| Econbrowser | market_news_alternative |
| A Wealth of Common Sense | market_news_alternative |
| The Big Picture (Ritholtz) | market_news_alternative |
| Farnam Street | market_news_alternative |
| Asymco | market_news_alternative |
| The Daily Upside | market_news |
| Rest of World | emerging_markets |
| TechCabal | emerging_markets_africa |
| Techpoint Africa | emerging_markets_africa |
| How We Made It In Africa | emerging_markets_africa |
YouTube RSS (1)
| Source | Category |
|---|---|
| Y Combinator | startups_ipo |
Full machine-readable list: GET /v1/knowledge/sources
Pricing
| Tier | Price | Rate limit | Features |
|---|---|---|---|
| Trial | Free | 100 req/day | Latest items, 3 days |
| Pro | $20/month | 1,000 req/day | Latest + semantic search |
| Fleet | $99/month | 10,000 req/day | Everything, bulk access |
Upgrade via /v1/payments/checkout (Stripe or crypto).
Self-hosting
git clone https://github.com/AgentBC9000/agentdb
cd agentdb
cp .env.example .env # fill in ANTHROPIC_API_KEY, DATABASE_URL, ADMIN_SECRET
docker-compose up
The scraper runs independently — trigger it manually or set a cron:
cd scraper
ANTHROPIC_API_KEY=... AGENTDB_API_URL=... ADMIN_SECRET=... python run.py
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。