Skip to main content
Managed Research is the easiest path to run MCP-driven optimization loops against your codebase. Connect a repo, define your research goal, and run overnight experiments in a managed environment. Runs are started by MCP. The public OSS client and MCP package lives at synth-laboratories/managed-research and is published on PyPI as managed-research.

​
Supported interface

Managed Research is MCP-first and MCP-only for external integrations right now.
  • Use the Managed Research MCP server/tools for automation and agent workflows.
  • UI remains supported for interactive usage.
  • Direct REST usage for SMR is considered internal/unstable and is not the supported external integration path.

​
Get started in one flow

  1. Configure MCP
  2. Create a project with smr_create_project
  3. Add repos and starting data
  4. Trigger a run with smr_trigger_run
  5. Pull artifacts with smr_list_artifacts

MCP Setup

Install the Managed Research MCP server, configure auth, and run a smoke test.

Quickstart

Create a project, complete onboarding, and trigger your first run.

Concepts

Runs, tasks, artifacts, budgets, approvals, and the harness — explained.