Automate the busy work. Keep the decisions.
rterminal takes the grind off your desk — searching, reading, extracting, wiring things up, and running long workflows end-to-end — and moves at a pace you can't by hand. Every judgment call and interpretation stays with you.
£20/month · 7-day free trial · cancel anytime
Runs on macOS & Linux · built on the open-source VS Code core.
One terminal. Every agent, every kind of output.
Pick a backend per session — switching never changes how you work — and let it answer in whatever form the task needs. Beside the chat, a display window fills with the artifacts it produces, each on its own tab.
Big workflows run for hours. You stay at the wheel.
rterminal will scope a question, survey the literature, run a hypothesis tournament, design and run experiments, and draft the paper — across many turns. It doesn't run off on its own: at every fork it hands the decision and interpretation back to you.
- Search, read & screen hundreds of sources
- Extract, triangulate & build the evidence ledger
- Run experiments & record measured results
- Draft, cite & check the write-up
- Confirm the scope & premises
- Pick the direction; sign off the design
- Interpret results; rule on each hypothesis
- Approve what ships
Ready for research out of the box. General enough for anything.
A working research method comes loaded — deep research, literature review, hypothesis tournaments, experiments, a paper pipeline. But underneath it's a general agent terminal: point it at a codebase, a dataset, an ops task, or a pile of writing, and it works the same way. The skills are just editable instructions — change them, add your own, and it adapts to how you work.
- ✓ Skills are plain, editable files — tune the method or write new ones
- ✓ Not just research: code, data, analysis, writing — any agent task
- ✓ Approval modes from read-only to full access, per session
Designing a Bayesian decision-support system for sentencing.
A live session — scoped, reviewed, and decided with the researcher. The busy work below was automated; the direction and the verdicts were the human's.
284 sources screened down to 18 load-bearing claims, each traced to its sources.
Candidate designs, scored by an automated review panel — each kept or revised, with reasons.
- 1 Integrated system 1265
- 2 Analogical — adapt a bail cost-min model 1241
- 3 Lean minimal-sufficient champion 1231
Before anything shipped, the run checked itself: references resolve, verdicts cover every hypothesis.
Sign in once. Everything follows you.
Your skills and tools live in your account and sync to every machine you sign in on — no settings to touch. A new account starts from a curated template you can edit and grow, and your library stays yours: an independent copy, access-controlled and audited.
- ✓ Skills & tools sync down on sign-in and back up as you edit
- ✓ Runs as ordinary local mode — nothing skill-aware on the server
- ✓ Tokens stored hashed, fetches rate-limited, access audited
Put the busy work on autopilot.
Start free for 7 days, then £20/month. Sign in from the app and your skills are set up for you — no configuration.