Build with Korely
A memory service for AI agents. Your agent saves text, Korely turns it into dated facts, and your agent reads them back when it needs them. Your data is stored on Korely's servers in Helsinki. The project's region decides which model reads your text to write the facts: global (the default) uses a model from Google, outside the EU; eu uses Scaleway in Paris.
Platform
What Korely does, how the pieces fit together, and what runs where.
korely.add("prefers Rust", user_id="dana") korely.get_context(query="stack?", user_id="dana")
Core operations
Add, search, get context, update and delete: the calls every agent makes, with code.
The graph & temporal facts
Facts name the people and things they are about, and carry the dates they were true. Reads return the facts that are true now, unless you ask for the past.
Cookbooks
Worked examples: a chatbot that remembers, a patient assistant, and more.
Integrations
Use Korely from LangGraph, the Vercel AI SDK, n8n and other agent tools.
API Reference
Every REST endpoint, with request examples and the shape of each response.
Where to start
New to agent memory? Read the Quickstart: it gets you a free key and stores and reads back a first memory in a few minutes. The words used across these docs, such as memory, fact and end user, are defined once in the Glossary.
- You write code: Quickstart, then the core operations.
- You want an agent to get its own key without a form: Sign up as an agent.
- You use a coding assistant such as Claude Code or Cursor: Vibecoding and Korely MCP.
- You are coming from another memory tool: the migration guides.
- You have a question about plans, data or limits: the FAQs.