Raycaster

Raycaster is an evaluation and benchmarking company for AI agents doing consequential document, spreadsheet and tool based work. Its public benchmark site publishes each task, the agent's execution trace, the output files and a rubric scored verdict, so a buyer can inspect exactly what an agent did before trusting it with similar work. The stated goal is to make agent work observable. It serves organizations in regulated and high stakes fields, including biopharma (CMC, quality, clinical and submission documents), legal (contracts, diligence packets and exhibits), healthcare records, energy and EPC submittals, finance and commercial operations. Three services sit behind the benchmarks. Evaluations are domain specific tasks with rubrics and reproducible verdicts, built with practicing experts who turn professional standards into acceptance criteria. Evaluation data and environments provide task sets, graders and reproducible workspaces for training, licensing and regression testing. Reference applications, Workspace and Sheets, run the workflows themselves: Workspace reads DOCX, XLSX, PDF and PPTX files, completes the assignment inside the files while preserving formulas and citations, and lets a reviewer inspect every change and the agent trace before accepting the result. Repeatable tasks can be saved as reusable agents. Published suites include Biopharma Bench, SpreadsheetBench and OfficeQA Pro. Pricing is not published. Access begins with a project brief or a scoping call, after which Raycaster returns a scoped proposal, relevant examples or a direct answer. The published benchmark results on the eval site can be browsed without an account.