Klarety SDK + CLI
from klarety import Nexus nexus = Nexus.connect() # Any file in, one table outnexus.ingest("./orders.csv", into="orders") totals = nexus.sql("SELECT region, sum(total) FROM orders GROUP BY region") # How does a supplier reach a customer?route = nexus.knowledge_graph.path(("supplier", "Nordvik AS"), ("customer", "Halden Foods")) # Changed your mind? Put the table backnexus.tables.rollback("orders", snapshot_id=8814205730)NORMALmainorders.py1:1
term://klarety
Everything a builder needs. One API, every way in.
Python SDK
Typed, with errors you can catch. Everything the web app does.
Command line
Tables, files, SQL and agents from your terminal.
REST API
Every route documented in one OpenAPI spec.
MCP for your AI editor
The same API, as tools your AI editor can call.
Spark when you need it
A real Spark session, one call away.
API keys you control
Keys that expire. Browser keys can only read.
Open tables
Read your data from any engine that speaks Apache Iceberg.
Web app, SDK, command line or your AI editor. The same data and rules behind each one.
One platform. Every part stands on the same data.
- NexusEvery record in one place that cleans, checks and keeps every version.
- StudioAI builds apps, agents and models on your data, and asks before big changes.
- AgentsAutomations described in plain words that run on a schedule or the moment data moves.
- BuddyReviews every change the AI makes, suggests what to build next, and fixes it in one click.
- ConnectionsConnect any app, API, database or bucket by describing it in plain words.
- ModelsFind any model on Hugging Face, then use, fine-tune or host it by describing what you need.
- Knowledge graphCustomers, orders, suppliers and documents, linked into one map.
- SandboxAI builds on real data in a sealed space of its own, and asks before big changes.
- DeployShip from GitHub with private previews, scaling and one-click rollback.
- AI harnessBring Claude Code, Codex or Cursor into a shared workspace, each person on their own account.