ORCH — CLI runtime for parallel research agent teams with state machine + review gates

[oxgeneral]

Hey gpt-researcher community — sharing a tool for orchestrating parallel research pipelines.

The problem: GPT-Researcher does an excellent job running a deep research workflow end-to-end, but when you need multiple specialized research agents running in parallel — e.g. one agent researches market data, another competitive intel, a third synthesizes sources — you end up manually coordinating: routing outputs, tracking completion, handling crashes, combining results.

ORCH provides the coordination infrastructure: a CLI runtime with a validated task state machine (todo → in_progress → review → done), auto-retry when an agent stalls, inter-agent messaging (agents share findings via a shared context store), and isolation via git worktrees. Each GPT-Researcher instance runs as an ORCH agent with its own scope.

npm install -g @oxgeneral/orch
orch agent add "Market-Researcher" --adapter shell --role "Research market trends"
orch agent add "Competitor-Researcher" --adapter shell --role "Research competitors"
orch agent add "Synthesizer" --adapter claude --role "Combine research into report"
orch task add "Market analysis Q1" --assignee Market-Researcher
orch task add "Competitor landscape" --assignee Competitor-Researcher
orch run --all

Parallel execution, automatic retry on failure, shared context for synthesis agents.

GitHub (MIT, 1647 tests): https://github.com/oxgeneral/ORCH

How are others here managing parallel research tasks across multiple agent instances?