AskChem: Claim-Centered Infrastructure for Chemistry Literature Synthesis
AskChem is a claim-centered infrastructure for cross-paper chemistry search that converts papers into atomic, typed claims grounded by source DOI and verbatim quote or evidence locator. It indexes 2.4M claims from 147K papers and provides a web interface, REST, SDK, and MCP access for AI agents. On AskChem-Bench, grounding a GPT-5.5 reader in AskChem yields 100% resolvable DOIs, compared with 88.3% without.
AskChem introduces a claim-centered infrastructure for chemistry literature synthesis, shifting the retrieval unit from papers to provenance-carrying claims. Each paper is decomposed into atomic, typed claims with source DOI and verbatim quote or evidence locator. The system indexes 2.4M claims from 147K papers and offers a faceted taxonomy, an evidence graph, and a living taxonomy for search and synthesis. It provides web, REST, SDK, and MCP interfaces for both human scientists and AI agents. Benchmarking shows that a GPT-5.5 reader grounded in AskChem achieves 100% resolvable DOIs, outperforming the 88.3% baseline.
AskChem restructures chemistry literature retrieval around atomic, typed claims with explicit provenance (DOI and verbatim quote). It builds a shared claim store with complementary structures: a stabilized faceted taxonomy for hierarchical browsing, an evidence graph linking claims through relations, and an exploratory living taxonomy mapping papers to scientific principles. The system supports AI agent access via MCP, enabling programmatic cross-paper synthesis. The benchmark improvement to 100% DOI resolvability suggests that claim-level grounding significantly enhances factual accuracy in AI-assisted literature review.
AskChem represents a shift from document-centric to claim-centric scientific search, which could improve the efficiency and reliability of literature synthesis in chemistry and potentially other scientific domains. By providing structured, machine-readable claims with provenance, it enables AI agents to perform more accurate and verifiable cross-paper reasoning. The availability of SDK and MCP access indicates a focus on integration into automated research workflows, which may accelerate adoption in pharmaceutical and materials science industries.
AskChem can reduce the manual effort and error rate in chemistry literature synthesis, potentially accelerating research cycles in drug discovery, materials science, and chemical engineering. Its AI-agent-friendly interfaces (REST, SDK, MCP) enable integration into automated pipelines, offering value to enterprises seeking to leverage large language models for scientific knowledge management. The claim-centered model may also support compliance and reproducibility requirements by providing verifiable provenance.
Observable next signals include: expansion of the claim store to more papers and chemistry subfields; integration with commercial AI research assistants; adoption by pharmaceutical R&D teams for drug discovery literature review; extension of the claim-centered approach to other scientific disciplines; and development of benchmarks for claim-level retrieval in broader contexts. The 100% DOI resolvability result may drive further research into provenance-grounded AI systems.