Deep Orchestrator: A Simple MCP Loop That Makes Deep Research Work
Read ArticleRead Original Articleadded Sep 12, 2025September 12, 2025
Qadri rebuilt a deep research agent for mcp-agent through three iterations, discovering that simpler architectures beat complex adaptive workflows. The winning approach—Deep Orchestrator—loops plan, execute, and deterministic verify, adds selective external memory via task dependencies, uses a TODO queue with parallelism, structured prompts, and a minimal policy engine. It delivers reliable performance purely via MCP, with future work on remote execution, tool selection, memory as MCP resources, and dynamic model selection.
Key Points
- A simple plan–execute–verify loop with replanning outperforms complex adaptive systems for deep research workflows.
- Deterministic checks (dependency graph, server and agent existence) paired with LLM planning greatly reduce hallucinations and failures.
- Selective external memory propagation via planner-declared task dependencies improves token efficiency without losing needed context.
- Generate a full upfront plan to populate a TODO queue (with parallel subtasks) and use a small policy engine instead of elaborate mode detection.
- MCP-centric design enables general-purpose agents; future work targets remote execution, intelligent tool selection, memory as MCP resources, and dynamic model selection.
Sentiment
Largely positive and aligned with the article’s thesis that a strong reasoning model is critical for planning; discussion revolves around which top-tier model to choose.
In Agreement
- Use a large, strong reasoning model for the planning phase; plan quality heavily influences final results.
- Errors compound from a weak plan, reinforcing the article’s emphasis on high-quality upfront planning.
- Candidate planner models mentioned include Claude Opus 4.1, GPT-5, and Gemini 2.5 Pro; OP notes Claude Code and similar agents also prioritize strong planners.
Opposed
- No explicit opposing viewpoints were raised; one commenter wondered whether open-source reasoning models could perform well, indicating uncertainty rather than disagreement.