Skip to main content
Research Factory is for multi-run research programs: factories, efforts, and project benches that compose many Managed Research runs into one inspectable campaign. It uses the same public integration path as Managed Research:
  • Python: SynthClient().research
  • MCP: hosted Managed Research MCP (smr_* tools)
  • Web: /smr/factories for interactive review
There is not a separate package or docs tab. Install synth-ai[research], construct SynthClient, and use client.research for both one-off runs and factory-scoped work.

​
When to use Factory vs a single run

ShapeUse
One goal, one evidence bundleManaged Research run — Quickstart
Many runs under one factory or benchResearch Factory — start from a factory project, then inspect runs and artifacts together

​
Next steps

Managed Research intro

Core run loop, projects, evidence, and launch fields.

Data factory cookbook

Intake → generate → review → publish for dataset-oriented factory work.

SynthClient

client.research namespaces and configuration.

MCP reference

Tool list for agent-driven factory and run control.