What is Kit For AI
Kit For AI is a memory and knowledge platform for AI agents. It provides persistent memory, document conversion, and retrieval-augmented generation (RAG) capabilities through a single API and MCP (Model Context Protocol) server. The platform allows AI agents to remember user preferences, search knowledge bases, and cite sources without requiring users to build and maintain a separate vector database, embedding pipeline, or chunking infrastructure.
How to use Kit For AI
- Sign up at kitforai.com/register (free, no credit card required).
- Add documents by uploading files (PDF, Word, Excel, PowerPoint, CSV, HTML, images) or providing URLs (web pages, YouTube videos). The platform converts them to clean Markdown.
- Create knowledge bases by grouping documents into searchable collections.
- Connect your AI agent via MCP (one command install for Claude, Cursor, etc.) or REST API.
- Use memory tools – your agent can call
remember, recall, and search tools mid-conversation to persist and retrieve information.
Features of Kit For AI
- Memory API: Persistent memory with
remember, recall, and search tools. Auto-created Memory knowledge base, near-duplicate dedup, versioned memories.
- Document conversion: Convert PDF, Word, Excel, PowerPoint, CSV, HTML, images (OCR), and YouTube videos to clean Markdown or JSON. Batch up to 25 files, URL fetch, same-domain crawl, scheduled URL refresh.
- Knowledge bases: Group documents into searchable collections. Grounded chat with cited answers, feedback corrections that stick, visual graph of connected docs, @mention to scope.
- Semantic search: Hybrid retrieval combining vector embeddings and full-text search with Reciprocal Rank Fusion and reranking.
- MCP-native: Works with Claude, Cursor, and any MCP agent. Same tools available over REST.
- Token efficiency: Up to 90% fewer tokens by pulling only relevant passages instead of full documents.
- Privacy: Encrypted at rest, API keys hashed, spaces isolate projects, delete means gone.
Use Cases of Kit For AI
- AI agents with memory: Give your AI agent persistent memory so it remembers user preferences, past decisions, and context across sessions.
- Document Q&A: Upload manuals, reports, or research papers and let your AI answer questions with citations.
- Chat with YouTube videos: Paste a YouTube link and the transcript becomes a searchable knowledge base.
- Web content grounding: Convert any URL to clean Markdown for your AI to use.
- Reduce token costs: Replace full-document context dumps with targeted semantic search.
FAQ
What formats does Kit For AI support?
PDF, Word, Excel, PowerPoint, CSV, HTML, images (with OCR), YouTube videos, and web URLs.
How does memory work?
Your AI agent calls remember, recall, and search tools via MCP or REST. Memories are automatically stored, deduplicated, and versioned.
Is my data private?
Yes. Files are encrypted at rest, API keys are hashed, projects are isolated, and deletion removes all data permanently.
Do I need to set up a vector database?
No. Kit For AI handles chunking, embedding, and retrieval internally. No vector DB, no embedding pipelines, no chunking configuration needed.
What models does it work with?
It works with any LLM via MCP or REST, including OpenAI, Claude, Gemini, Meta, Mistral, Perplexity, DeepSeek, Cohere, Grok, and Qwen.
