Agentic Research Directory

Tool

Onhand: contextual research-paper reader

Onhand documents a Chromium side-panel AI reading assistant for research papers, webpages and PDFs. Its official guide describes passage highlighting, paper explanations, following references and comparing clearly related tabs; the runtime is browser-hosted, while model inference uses configured external services.

Workflow level n/a Source: Repository-documented Outcome: Not empirically evaluated here

Research applicability

A directly documented aid for reading and comparing academic papers in context, not merely a generic chatbot inferred from research metadata. Researchers must inspect original passages and citation targets and supply domain judgment; using it as an aid to literature synthesis is editorial guidance, not a validated systematic-review method.

Getting started

Editorial, untested suggestion: read the research-paper guide and privacy/permissions policy, then try the documented Chrome installation on one public, non-sensitive paper. Ask a bounded question and check every highlighted passage and opened reference against the original. Do not assume restored artifacts match a changed page.

Maintenance

Dated primary-documentation snapshot only; continued maintenance and release availability were not independently established. Mutable HEAD and the live website were not commit-pinned.

License

README declares Apache-2.0 and the retrieved LICENSE contains Apache License Version 2.0. Bundled pi packages, browser distribution and model-provider terms were not independently audited; this does not license third-party papers or authorize their upload.

Workflow & human oversight

No fixed automation level assigned: researcher-invoked contextual reading support, with documented tools that may navigate and inspect related sources. Researchers select material, authorize exposure and validate interpretations; no autonomous discovery or guaranteed truthful answers established.

Workflow level is not an evidence score; higher is not better evidence.

Approach

The README documents an in-extension runtime with extraction, PDF-search/read, citation-find, highlight and tab tools. The official guide describes answering about the open paper and consulting cited or clearly related papers, with highlighted evidence and margin notes. Saved sessions/artifacts use local extension storage, but model requests leave the device under the selected authentication/provider mode.

Limitations

Documentation-only review of the README, official paper-reading guide, repository privacy policy and LICENSE; no extension installation, browser run, provider call or source-code audit. Highlighted passages and cited answers are documented capabilities, not guarantees of answer truth, citation accuracy or complete paper coverage. Local session storage does not mean local inference: relevant page/PDF content, prompts, selected text and possible screenshots may leave the device, and related-tab titles/URLs and relevant content can be inspected. The policy describes direct provider calls or an Onhand Free Cloudflare Worker forwarding to OpenAI, with required free-tier diagnostics; these privacy claims were not independently traffic-audited. No universal full-document upload or no-upload guarantee is established. Powerful debugging/scripting permissions require trust. README says changed-page/missing-tab restore remains in progress; archived runtime plans and historical evaluations are not current capability guarantees. No research accuracy, performance or productivity outcome evaluated.

Reproducibility

Exact HTTPS response bytes, UTF-8 bodies, requested/resolved URLs, UTC access timestamps, headers, raw-body SHA-256 and envelope SHA-256 saved. No upstream software or claimed outcome reproduced.

Adoption requirements

Requires a compatible Chromium extension environment and explicit acceptance of powerful browser debugging/scripting permissions. The README documents free hosted usage with daily caps and required diagnostics, Codex sign-in, or supported provider API keys; provider access and terms vary. Review active and related tabs before use, close sensitive material, and obtain authorization before any confidential, licensed or identifying text/image is sent to providers or the hosted Worker. Local saved sessions can retain sensitive artifacts and credentials; review deletion and retention settings.

Comments

No comments yet.

Add a comment

Comments are held for moderation before appearing. Critical or corrective comments are welcome — especially evidence that an entry overstates what has been verified.

Submitting stores your comment text, the optional name and email, and a pseudonymised connection identifier used only for spam limiting (the raw IP address is not stored).