FrontierOps
Track research with the coding agent you already use. Compare each search with your previous run, inspect the evidence, and turn selected papers into clear briefings. Your configuration and run history stay in your workspace.
More than a paper list
A useful briefing explains the question, the proposed method, the evidence behind the results, and what the experiments do not establish. Claims can be traced back to their sources; a ranking score is not a quality verdict.
Discovery is manual. Nothing searches on a timer. A new run compares the current scope with previous results and keeps earlier runs available.
Four actions
- 01
Setup
Install the harness and initialize a research workspace.
- 02
Config
Describe a topic, author, venue, or paper; review the proposed settings.
- 03
Discover
Search on request, compare runs, and choose papers to explain.
- 04
Briefing
Verify source material and publish a per-paper explanation.
The agent drives the internal Build, Sync, Rank, Brief, Review, and Render stages. Each stage leaves inspectable files.
Read or present
Interactive HTML
Move between the question, method, results, and limitations; open supporting evidence and paper pages.
OpenSlide
Keep an editable slide project that can be built and presented in a browser.
Video
Optional HyperFrames output when video and voice dependencies are configured. A preview is not an exported MP4.
Start in your own workspace
Install an Agent Skill, create an active configuration, and run Discovery when you want to look for changes.