Research that
improves itself
Lumen searches the web, synthesises sources, writes structured articles — then critiques its own work and iterates until quality meets the bar.
Six nodes. One loop.
Each research run flows through a pipeline that plans, searches, summarises, outlines, drafts — then reflects and decides whether to accept or revise.
Articles That Improve Themselves
The agent reviews its own work, identifies gaps in evidence or structure, and loops back to fix them — automatically.
Watch Every Step Live
See the pipeline work in real-time — which sources it found, what outline it planned, and why it decided to loop back.
Research Any Domain
Switch between medical, legal, financial, or general research — each with domain-specific search, terminology, and quality checks.
Quality You Can Measure
Every article is scored on quality, relevance, and groundedness. Track how scores trend across runs.
Every run is scored
An LLM judge evaluates each article on quality, relevance, and groundedness. Regressions are flagged automatically so you catch them before users do.
Score Trends
| Topic | Quality | Relevance | Grounded | Date |
|---|---|---|---|---|
| AI agents in software development 2026 | 4.2 | 4.5 | 3.9 | Mar 20 |
| GLP-1 receptor agonists cardiovascular outcomes | 4.5 | 4.3 | 4.1 | Mar 20 |
| Google antitrust ruling Chrome divestiture | 4 | 4.2 | 3.8 | Mar 19 |
| Trump tariffs impact on global markets | 3.8 | 4.1 | 3.6 | Mar 19 |
Try it with your own topic
Bring your own API key. No usage caps on our end.