Data Workers

Data Workers, Inc. makes an agentic data platform built around specialised data agents and a living context graph. The company describes Data Factory 2.0 as four products: Data-Agents Swarm, Data Context Wizard, Autonomous Data-Conductor and Spellbook Data Catalog. An Apache-2.0 community repository ships the agents as MCP servers for coding agents such as Claude Code, Codex CLI, OpenCode, Gemini CLI and Cursor.

What It Is

Data Workers is an autonomous data-operations platform. Agents detect, diagnose, fix, review and verify data work across warehouses such as Snowflake, Databricks and BigQuery, with a named human in control and a receipt on each change. The homepage lists agents for incident debugging, pipeline building, streaming, schema evolution, MLOps, connectors, ingestion, observability, quality monitoring, change review, cost cleanup, usage intelligence, access governance, security and identity.

The Four Products

• Data-Agents Swarm: specialised agents that own work end to end, such as incident resolution, pipeline builds and change review, with governed write access.
• Data Context Wizard: a semantic and knowledge graph of the data layer that every agent reads from.
• Autonomous Data-Conductor: the orchestrator that runs the detect, diagnose, fix, review and verify loop and directs the swarm.
• Spellbook Data Catalog: the place to command, approve and audit the agent fleet.

Community Edition and Setup

The community repository contains 11 agents, 160+ MCP tools and 15 catalog connectors covering pipelines, incidents, catalog and lineage, schema, quality, governance, usage, observability and ML. It runs locally with in-memory seed data by default, and users can bring any LLM provider. Setup is via npx dw-claw init and adding the MCP server to a client, with an optional interactive step to connect Snowflake, BigQuery or Databricks. The README says write operations such as pipeline generation and model training require Pro, and the full platform is offered through dataworkers.io.

Vendor Claims

The homepage reports a 74.4% task success score on DataAgentBench (UC Berkeley EPIC Data Lab) and estimates of faster completion for routine data tasks. These figures are the vendor's own, with methodology in its white paper.