Get started

Quickstart

Give your coding agent memory in one command — local, zero infra. Then watch it grow into live data and a graph of your whole stack.

kyma ~ install

$curl -fsSL https://raw.githubusercontent.com/shakedaskayo/kyma/main/install.sh | bash

Start here

Local · zero infra · ~5 minGive your agent memoryInstall, wire it to your coding agent (plugin / CLI / MCP), and it recalls durable, graph-aware memory into every prompt — across sessions and machines. This is the fastest, most useful way in.kyma setup claude-codeStart here →

Go further

Five-minute start · run the full engine — docker compose, ingest, KQLHow kyma works · memory → data → graph, plus modes, dreaming, syncFirst real run · a batched ingest, a real summarize, the agent endpointConcepts cheatsheet · the invariants, default ports, every KYMA_* var

The shape of it

kyma grows in three steps — each one is the same engine and the same graph, doing more:

1. Memory — your agent remembers across sessions (Give your agent memory).
2. Live data — point logs, traces, and code at it; query in KQL/SQL (Five-minute start).
3. The graph — data sources build a property graph and memory links to the real resources it's about (How kyma works).

Where to go after

• The mental model, end to end: How kyma works · Concepts.
• More ways to get data in: Ingest · Data sources.
• More ways to ask questions: Query.
• What it's for: Use cases.