Find a research tool you can evaluate
Start with a research task, then decide what a useful result would look like. The directory helps you compare documented approaches; inclusion is not an endorsement or proof that a tool will improve your research.
This draft’s navigation and label descriptions were checked against saved local site files, not the live deployment; the live site may differ.
1. Write down one task
Choose something specific: finding papers for a question, screening a reading list, or understanding a passage. Describe the output you need and one check you can perform yourself.
For example: “I want a shortlist of papers about my question, with enough information to check why each paper belongs.” Treat this as a suggested starting point, not a tested workflow.
2. Explore, then narrow the results
Open Explore and search with your task or subject. Use the filters to narrow resource type, discipline, task, evidence or workflow level. If nothing matches, clear a filter or broaden the search; an empty result is not proof that no relevant tool exists.
Use a resource page to answer three different questions:
- What does it document? Read the resource description and its linked primary sources.
- What evidence supports the entry? Check source level, outcome verification and the date sources were checked.
- What does the human still do? Read the workflow and human-oversight explanation before deciding whether it fits your task.
3. Read the labels separately
Source level describes where an entry is grounded. Outcome verification describes how far its reported results have been checked. A peer-reviewed source can still describe author-reported results; source checking does not establish independent reproduction.
Workflow levels describe the scope of the documented use. A higher level is not a quality ranking, and an unclassified entry is not automatically unsuitable. See Method for the definitions and each entry for its qualifications.
4. Compare a small set of examples
These entries illustrate different tasks rather than a ranked shortlist:
| Your question | Example to inspect | Check before choosing |
|---|---|---|
| How is an agent-assisted literature review organised? | Priorwork | Does its documented process match your review question and the decisions you need to retain? |
| What is documented about literature screening? | LatteReview | Which screening decisions would you check yourself, and against what criteria? |
| What is documented about contextual paper reading? | Onhand | Can you trace an explanation back to the paper passage and its surrounding context? |
Open the resource’s primary source before installing software or following setup instructions. These examples were recorded as published in the saved catalogue on 19 October 2026 at 08:15 UTC; their current pages, availability and capabilities were not rechecked for this draft.
5. Plan one small, checkable trial
Use public or synthetic material for the first trial. Check upload, retention and reuse terms before introducing confidential or restricted material.
Keep a short record:
- Your question, input material and expected output.
- The resource, version and settings you used.
- The output, plus the source passages or records you checked against it.
- What was wrong, missing or unsupported, and your decision to continue, revise or stop.
For a paper shortlist, check that the papers and identifiers exist and that the papers actually support the stated reasons for inclusion. Keep the final inclusion decision with the researcher. These are editorial suggestions, not a reproduced result or a promise of time savings.
For more planning examples, open Research Playbooks. They connect documented capabilities with suggested workflows; they do not certify privacy, rank tools or reproduce research outcomes.
Local proposal for review. No installation, upload, tool trial or site publication was performed in preparing it. The accompanying internal EVIDENCE.json records the source files, hashes and claim locators.