Consensus Accelerates Research with GPT-5 and the Responses API

Research assistant platform Consensus has unveiled Scholar Agent, a multi-agent system built on GPT-5 and the Responses API. Going beyond simple search, this system dramatically speeds up research by mimicking the process researchers use to go from a question to a conclusion.

Scholar Agent consists of four specialized agents that perform precise reasoning:

• Planning Agent: Analyzes the question and decides the next action
• Search Agent: Searches the paper index, personal library, and citation graph
• Reading Agent: Interprets papers individually or in batches
• Analysis Agent: Synthesizes results, decides on structure and visualization, and writes the final output

Consensus adopted a context engineering approach to increase reliability. This method, which gathers relevant evidence before generating an answer, minimizes hallucinations by providing every answer with a 'research context pack' that includes papers, metadata, and key findings.

The company also switched from the existing Chat Completions to the Responses API to support multi-agent routing. This improved cost efficiency and control, and by leveraging GPT-5's strong long-context reasoning and tool-calling capabilities, Consensus confirmed performance superior to GPT-4.1, Sonnet 4, and Gemini 2.5 Pro.

Rather than an institution-focused sales approach, Consensus grew rapidly by directly targeting individual researchers such as students, professors, and clinicians. Recently, it has expanded into specialized domains by signing a contract with Mayo Clinic's medical library and launching Medical Mode for medical professionals.