Agents

The researcher agent searches, reads, and extracts findings from papers and web sources.

The reviewer agent evaluates documents with severity-graded academic feedback.

The writer agent produces structured academic prose from research findings.

The verifier agent cross-checks claims against their cited sources.

The hypothesis agent converts evidence into falsifiable, test-ready hypotheses.

The contradiction agent stress-tests conclusions with counter-evidence and failure cases.

The evidence scorer agent quantifies support strength for each claim.

The experiment agent designs and executes minimal tests to validate key hypotheses.

The citation integrity agent verifies that each citation supports the exact claim text.

The memory agent maintains durable project memory across sessions.

The orchestration brain that decides agent sequence, checkpoints, and stop conditions.

The knowledge graph agent structures findings into entities and relationships.

The reasoning validator checks whether conclusions logically follow from evidence.

The iteration controller runs multi-loop research cycles to convergence.

Collects candidate papers with structured metadata for review pipelines.

Scores and filters literature candidates by relevance, influence, and recency.

Groups selected papers into themes and summarizes each thematic cluster.

Identifies missing areas, underexplored settings, and high-value open problems.

Produces survey-style literature reviews with themes, disagreements, and gaps.

Formalizes equations, notation, and derivations for technical paper sections.

Produces publication-ready figure/table plans with traceable captions and data paths.

Designs manuscript structure, claim map, evidence map, and reviewer-risk framing before drafting begins.