Deep Research With Agents and Cited Sources
Agentic AI research in Rivul AI: ask a question in the Research notebook for a cited brief, or run a research agent that searches the Rivul Index and your library, reads the most relevant papers, and writes a cited report.
Free for every account. Search needs no key; briefs and agent runs use your own AI provider key (BYOK).
What Deep Research Means Here
Agentic AI research becomes useful when it can do more than generate prose: it must decide what to search, gather relevant scholarly records, compare evidence, expose uncertainty, and show the researcher where every important claim came from. Rivul AI keeps that chain visible in the notebook and in Runs.
Search, Investigate, Verify, and Write
Search through Rivul Index
Search metadata for 300M+ academic and scientific papers, then keep the title, authors, year, DOI, venue, and source location of each record the research used.
Run an agent in steps
A research agent plans its own searches, reads the most relevant papers, and writes a cited report. Quick runs take up to 5 model steps and Thorough runs up to 12, all on your own AI key.
Keep evidence beside the answer
The Research notebook lists every source it found as Cited or Not used, and Runs keeps every step an agent took. Open the paper before accepting a synthesis.
Draft with traceable citations
Carry reviewed findings into a paper with their source records attached, then check claims, manage the bibliography, and export the manuscript.
Research Notebook, Agents, Rivul MCP, and the Index
Research notebook
A question becomes a cited brief and a findings table from the Rivul Index and your Library, with every source listed as Cited or Not used.
How Research worksAgents and Runs
Saved research briefs with a goal, sources, and an output. Each run searches, reads, and writes a cited report, and Runs keeps its steps and sources.
How agents workRivul MCP
The MCP server and REST API that let Claude Code, Codex, Cursor, and your own software search scholarly records and return citation metadata with source links.
Explore Rivul MCPRivul Index
The private scholarly search and metadata engine beneath Research, agents, and Rivul MCP. It returns records and source locations; it does not turn metadata into proof by itself.
Review data sourcesWhat the Researcher Still Decides
A source-linked workflow can reduce repetitive searching and organization. It cannot decide whether a study is valid, whether a method fits your question, or whether a passage supports your final wording. Rivul AI keeps those decisions visible instead of hiding them behind a finished answer.
Deep Research FAQ
What is deep research in Rivul AI?
Deep research in Rivul AI is multi-step research with a visible trail. The Research notebook turns a question into a cited brief and a findings table, and research agents run a saved brief that searches, reads, and writes a cited report.
What makes it agentic?
A research agent decides its own next step: it forms searches, reviews candidate papers, reads the most relevant ones, and writes a report, within a fixed step and token budget on your own AI key. Every step, source, and report is kept in Runs so the researcher can check it.
How does Rivul Index support deep research?
Rivul Index is Rivul AI's private scholarly search and metadata engine over 300M+ works. It returns source-linked records that can be inspected before a paper is used as evidence, and it needs no AI key.
Can I run deep research from Claude, Cursor, or Codex?
Rivul MCP gives MCP clients read-only scholarly search and paper records from the Rivul Index, so the client's own agent can search papers. Research agents with Runs work inside Rivul AI.
Does deep research replace reading research papers?
No. AI can help locate, organize, and compare evidence, but the researcher must read the relevant source, evaluate its methods and limitations, and confirm that it supports the final wording.