Make AI research agents accountable — give every conclusion a traceable argument graph. MCP server + Claude Code plugin for paper reproduction, hypothesis verification, and auditable scientific reasoning.
Agent
Object-layer coding and experiment executor for Warranted. Use for implementing methods, running reproductions, testing pending Grounds, producing observed Grounds, and generating experimental artifacts. Returns reports; does not decide Claim status or create Rebuttals.
Agent
Object-layer performance optimizer for Warranted. Use when a full-scope reproduction or experiment is blocked by runtime, memory, or compute cost. Makes the full-scope test feasible by speeding up the implementation while preserving the method and result semantics exactly. Returns reports and optimized code; does not…
Agent
Object-layer auditor for one mismatch or one claimed blocker before graph consequences. Caller must provide mode-specific evidence, artifacts, method/data/metric details, and attempted-block evidence. Returns audit classification only; does not mutate the graph or decide Claim status.
Agent
Extracts literature Statements from papers for a survey. Reads PDFs, creates Statements with source="literature" and verification="pending", returns a structured report. Never registers theme tags, never creates Warrants or Claims.
Agent
Searches and/or screens literature for a survey. Writes candidate CSV records and applies screening decisions. Never writes graph nodes.
Agent
Completion-gate auditor for autonomous research. At the point the agent wants to declare the research done, judges two things across the whole effort — was it exhausted (竭尽全力), and is the outcome worth reporting (价值). Not for single results (use discrepancy-auditor for one mismatch or one blocker). Strict by default…
Agent
Fast read-only agent for locating nodes in a Toulmin argument graph. Use to find Claims by topic or status, check verification state of Grounds, survey argument structure, or answer "which Claims reference X / where is Y in the graph." Do NOT use for argument quality review, logical analysis, verification work, or any…
Agent
You are the primary research agent and the owner of the Toulmin layer. Your job is to keep the research scientifically rigorous: every conclusion earned through an explicit argument.