Give Your AI Agent a Brain: Persistent Memory in 5 Minutes

Most AI agents are goldfish. They process a request, respond, and forget everything. The next conversation starts from zero.

This is fine for chatbots. It's a disaster for agents that need to know things — your documentation, your product specs, your team's decisions.

AgentBooks fixes this. It's a knowledge layer that gives your AI agent persistent, searchable memory with three API calls.

The Problem

You're building an AI agent with LangChain, CrewAI, or your own framework. Your agent needs to:

• Remember past conversations and decisions
• Search through your documentation
• Share knowledge with users via a readable page

The typical solution? Build a RAG pipeline from scratch:

1. Set up a vector database (Pinecone, Weaviate, Qdrant)
2. Write chunking logic
3. Build an embedding pipeline
4. Create a retrieval layer
5. Hope it all works together

That's weeks of infrastructure work before your agent can remember anything.

The AgentBooks Way

Step 1: Create a Space (30 seconds)

Sign up at dash.agentbooks.net, create a knowledge space, and grab your API key. Done.

Step 2: Upload Knowledge

curl -X POST https://api.agentbooks.net/api/v1/documents \
  -H "Authorization: Bearer ab_your_key" \
  -H "Content-Type: application/json" \
  -d '{
    "title": "Deployment Guide",
    "content": "# How to deploy\n\nStep 1: Clone the repo...",
    "tags": ["devops", "guide"]
  }'

AgentBooks automatically chunks your content and creates vector embeddings. No configuration needed.

Step 3: Your Agent Searches

curl "https://api.agentbooks.net/api/v1/search?q=how+to+deploy&limit=3" \
  -H "Authorization: Bearer ab_your_key"

Response in ~17ms:

{
  "results": [
    {
      "title": "Deployment Guide",
      "chunk": "Step 1: Clone the repository...",
      "score": 0.92
    }
  ]
}

That's it. Your agent now has persistent, searchable memory.

What Makes AgentBooks Different

The "Dual-Sided" Advantage

Most tools solve one side of the problem:

Tool	Machine API	Human Reading Page
Pinecone	✅	❌
Weaviate	✅	❌
Notion API	⚠️	✅
GitBook	❌	✅
AgentBooks	✅	✅

AgentBooks is dual-sided: every document your agent uploads is automatically available as a beautiful, SEO-friendly reading page for humans. Your agent writes knowledge; your users read it.

MCP Protocol Support

AgentBooks speaks MCP (Model Context Protocol) natively. Connect it to Claude Desktop, Cursor, or any MCP-compatible client:

{
  "mcpServers": {
    "agentbooks": {
      "url": "https://api.agentbooks.net/api/mcp",
      "headers": {
        "Authorization": "Bearer ab_your_key"
      }
    }
  }
}

Your AI assistant can now search and upload to your knowledge base directly.

Built-in Features You'd Otherwise Build Yourself

• Automatic chunking — No need to split documents manually
• Vector embeddings — Semantic search out of the box
• Tag filtering — Organize and filter by custom taxonomies
• API key scoping — Per-space keys with usage tracking
• File uploads — Images, PDFs, videos up to 20MB
• Custom domains — Each space can have its own domain
• Public or private — Control visibility per document

Real-World Use Cases

1. Documentation Agent — Upload your docs, let your support agent search and answer questions.

2. Research Assistant — Your agent collects and indexes research papers, then retrieves relevant findings on demand.

3. Team Knowledge Base — Agents and humans share the same knowledge. Agents write summaries; humans read them on a branded page.

4. Content Pipeline — Your content agent writes articles via API; they're instantly readable at your custom domain.

Get Started

AgentBooks is in public beta — free, no credit card required.

1. Sign up
2. Create a space
3. Start uploading

Your agent's memory starts now.

---

AgentBooks — The Knowledge Layer for AI Agents www.agentbooks.net · API Docs · hello@agentbooks.net