# Alkera

> Agentic data platform where humans and AI agents collaborate on data engineering, analysis, and science in shared workspaces.

Alkera AI, Inc. builds Alkera, an agentic platform for data engineering, analysis, and data science in collaborative workspaces shared by people and agents. It connects to warehouses, transformation tools, and knowledge sources, and offers notebooks, dashboards, column-level lineage, and automated pipeline maintenance. An open-source, self-hostable edition called Databench is published under Apache-2.0.

## What It Is

Alkera is a data-stack platform in which an AI agent works alongside team members in multiplayer chats and notebooks. In the homepage demo, the agent ingests market data into Snowflake, prepares it with dbt, extracts structured LLM signals with Python, models price trends, and builds a mart served to BigQuery, with human approval before publishing.

## How It Works

- **Knowledge base:** the agent reads context from connected knowledge sources, and each entry shows its sources and whether a human has verified it.
- **Column-level lineage:** lineage is traced across warehouse, transformation, and analysis, and agents check downstream impact before making changes.
- **Automated maintenance:** the platform detects job failures, stale data, and correctness issues. Agents propose fixes and test them in isolated sandboxes, and CI/CD gates enforce quality standards.
- **Notebooks and dashboards:** SQL and Python notebooks created with the agent can be shared as interactive dashboards, recalculating only the dependent results when an input changes.
- **GPU compute:** a notebook kernel can be given a dedicated GPU node for training and post-training.

## Enterprise and Deployment

The enterprise page describes deployment on your own infrastructure, with SSO/SAML, SCIM provisioning, team- and role-based IAM, per-member budgets, and an organization audit log. Access to data uses each member's own identity so existing data permissions still apply.

## Databench Open Source Edition

Databench is a self-hostable, web-based workspace with shared files, SQL and Python notebooks, agent chats, and compute on infrastructure you control. It runs through Docker Compose with an Anthropic or OpenAI API key. Chats run in sandboxes on local Docker or SSH-reachable Linux machines. It includes a web app and an `alkera` CLI, and the README describes Alkera as the commercial offering with more features.

## Features
- Shared multiplayer workspaces for humans and agents
- Knowledge base with human-verified entries
- Column-level lineage
- Automated pipeline maintenance with sandbox testing
- SQL and Python notebooks
- Interactive dashboards
- Dedicated GPU notebook kernels
- SSO/SAML and SCIM
- Team and role-based IAM
- Audit logs and member budgets
- Self-hosted deployment
- Open-source Databench edition

## Integrations
Airflow, dbt, ClickHouse, Databricks, DuckDB, Google Docs, Linear, MySQL, PostgreSQL, Sigma, Snowflake, Tableau, AWS S3, BigQuery, Confluence, Datadog, Fivetran, GitHub, Hex, Looker, Notion, Redshift, Slack, SQLite, Trino

## Platforms
WINDOWS, MACOS, LINUX, WEB, API, CLI

## Pricing
Freemium — Free tier available with paid upgrades

## Links
- Website: https://alkera.ai
- Documentation: https://docs.alkera.ai/
- Repository: https://github.com/AlkeraAI/Databench
- EveryDev.ai: https://www.everydev.ai/tools/alkera