Salesforce Enterprise Deep Research (github.com)

Salesforce Research released Enterprise Deep Research (EDR), an open-source, steerable multi-agent system for automating enterprise-scale research and analytics. EDR combines a Master Planning Agent for adaptive query decomposition, four specialized search agents (general web, academic, GitHub, LinkedIn), a Visualization Agent, and an extensible MCP-based tool ecosystem (NL2SQL, file analysis, enterprise workflow hooks). It adds a reflection loop that detects knowledge gaps and updates research direction, supports real-time steering and optional human-in-the-loop guidance, and streams progress via a web UI and Slack agent. The code, demos, and deployment stack (FastAPI backend + React frontend) are available under Apache 2.0 on GitHub. Technically significant for AI/ML practitioners, EDR integrates multi-provider LLM support (OpenAI, Anthropic, Groq, SambaNova, Google Gemini) with model-specific defaults and parallel processing for large-scale, iterative research. It ships an evaluation suite (DeepResearchBench, ResearchQA, DeepConsult) and the EDR-200 dataset of 201 complete agentic research trajectories, exposing the full planning/search/reflection pipeline for fine-grained benchmarking of agent behavior. Enterprise features include citation management, document analysis, NL2SQL, streaming results, and easy deployment (Python 3.11+/Node 20.9). EDR offers a reproducible platform to build, evaluate, and steer agentic research workflows in production settings.