{"slug": "show-hn-a-possible-open-source-desktop-alternative-to-openai-codex", "title": "Show HN: A possible open-source desktop alternative to OpenAI Codex", "summary": "An open-source desktop alternative to OpenAI Codex, called Open Science Desktop, has been released for macOS, Windows, and Linux. The tool provides a local-first, model-agnostic AI research workbench that connects agents, notebooks, files, and reports into an auditable workflow, and it ranks first on the ResearchClawBench benchmark for autonomous scientific research agents.", "body_md": "**Local-first, model-agnostic AI research workbench for macOS, Windows & Linux.**\n\nFormerly Open Science. An open-source desktop alternative to Claude Science and similar AI-for-science workbenches — built with Tauri, MCP, agent skills, and reproducible artifacts. It connects agents, notebooks, files, figures, reports, runs, and review into one auditable desktop workflow.\n\n🎉 **Recognition:** Open Science Desktop ranks #1 by scored-task average on [ResearchClawBench](https://internscience.github.io/ResearchClawBench-Home/), an end-to-end benchmark for autonomous scientific research agents (Pass@1 leaderboard, July 9, 2026).\n\n[✨ What it does](#what-it-does)[🎬 See it in action](#see-it-in-action)[🧪 Current capabilities](#current-capabilities)[🔌 Skills and connectors](#skills-and-connectors)[📦 Install](#install)[🚀 Build from source](#build-from-source)[🔒 Safety and privacy](#safety-and-privacy)[🗂️ Repository layout](#repository-layout)[📌 Status](#status)[🤝 Contributing](#contributing)[⚖️ License](#license)\n\n**Runs the whole research loop** — from a broad direction to a finished paper:\nexploration, literature survey, hypothesis, experiment code, analysis, figures, and\nwrite-up, in one continuous, auditable session.\n\n**Autonomous research agents**— the bundled`ai4s-agent`\n\nchains specialist skills end to end (explore → survey → experiment → write), and each stage drops a real, inspectable artifact into your workspace, not just a chat reply.**Everything traces back**— figures, tables, reports, notebooks, and run outputs link to the exact code, inputs, environment, model output, and conversation that produced them.**Local-first and yours**— sessions, data, provenance, notebooks, and run records live in local folders on your machine. Nothing leaves by default.**Model-agnostic runtime**— the UI talks through`packages/sdk`\n\nto a bundled, pinned OpenCode sidecar. Bring your own model; providers, skills, and MCP servers stay pluggable.**Reproducible by construction**— local, SSH/Slurm, Modal, and notebook-batch runs are captured as reproducible run records, not loose terminal scrollback.**Extensible**— agent skills, MCP servers and one-click science connectors,`/`\n\ncommands,`!`\n\nshell mode, and a model-agnostic SDK.\n\n**One prompt -> a complete, traceable analysis.** Simulate data, fit a model, save a\npublication-grade figure, and write a report where every number traces to the code.\n\n**Every artifact traces back to its code, inputs, and conversation.**\n\n**Literature -> verifiable report.** Search papers, draft a manuscript rendered as a\nPDF, and audit citations, unsourced numbers, and figure/code consistency.\n\n**The research loop, as skills.** One meta-skill runs the full pipeline; each stage\nis a self-contained skill that produces a real, gradeable artifact — runnable on any\nmodel OpenCode supports:\n\n| Skill | Role | Primary output |\n|---|---|---|\n`ai4s-agent` |\nRuns the four skills below, in order | The full research package |\n`research-explorer` |\nTurn a broad direction into concrete topics | `research_exploration.md` , `topic_matrix.md` , `literature_pre_survey.md` |\n`literature-survey` |\nWrite a literature survey | 6–20 pp PDF, 60+ real citations, LaTeX source, taxonomy figures |\n`experiment-suite` |\nBuild an experiment package | Design doc, runnable code, `results.json` with provenance, figures, report |\n`paper-writer` |\nWrite a research paper | 8–14 pp PDF, 200+ citations, 4–8 figures, tables |\n`mindmap-render` |\nRender a mindmap | Image generated from a `topic_matrix.md` |\n`integrity-auditor` |\nAudit a paper's integrity | Image / numerical / logical findings, 4-level evidence grading, `audit_report.md` |\n\nThese ship in the `ai4s-skills`\n\npack alongside first-party review skills and the\noffice/document skills below.\n\n| Area | Current state |\n|---|---|\n| Desktop shell | Tauri 2 + React + TypeScript + Vite, with macOS, Windows, and Linux desktop builds. |\n| Runtime | Bundled OpenCode sidecar, auto-started by the app, isolated from the user's own OpenCode config/data. |\n| Sessions | Multi-session chat/history, dated workspace folders, global history across workspaces, `/` commands, and `!` shell mode. |\n| Files | Global and per-session file browsing, context menu actions, external open/reveal, copy path, and local preview server. |\n| Notebooks | Real `.ipynb` files, Python and R notebook creation, local kernel execution, managed Jupyter environment via bundled `uv` , and an Open JupyterLab action. |\n| Runs | Append-only run logs, global SQLite run index, search/facets/pagination, local/remote surfaces, output links, logs, and reproduce prompts. |\n| Provenance | `.openscience/provenance.jsonl` tracks file versions and links produced artifacts back to the run or edit that created them. |\n| Review | Traceability, statistics-integrity, domain-check, large-file, publication-figure, remote-compute, and Modal run skills are bundled as first-party skills. |\n| Viewers | PDF, image, video, HTML, Markdown, code, CSV/TSV tables with charts, DOCX, XLSX, PPTX, molecules, 3D meshes, genome tracks, FITS, DOS/DOSCAR, EIGENVAL bands, qcode, anomaly maps, and phase files. |\n| Models | OpenCode provider catalog, OAuth/API-key provider flows, custom OpenAI-compatible endpoints, and local/provider-specific options supported by OpenCode. |\n| Interface languages | English, Simplified Chinese, Japanese, Spanish, German, French, and Korean. Portuguese (Brazil) and Arabic are registered but not selectable yet. |\n\nBundled skills are fetched for builds and releases instead of being committed into git history:\n\n`ai4s-skills`\n\npack from`ai4s-research/ai4s-skills`\n\n.- Office/document skills from the Apache-2.0\n`anthropics/skills`\n\nrepository:`docx`\n\n,`pdf`\n\n,`pptx`\n\n, and`xlsx`\n\n. - First-party core skills in\n`runtime/skills/core/`\n\n:`traceability-review`\n\n,`stats-integrity`\n\n,`domain-check`\n\n,`large-file`\n\n,`publication-figures`\n\n,`remote-compute`\n\n, and`modal-run`\n\n.\n\nOne-click science MCP connectors currently include:\n\n- Literature search: arXiv, PubMed, Crossref, Semantic Scholar, bioRxiv/medRxiv.\n- Biomedical databases: PubMed, ClinicalTrials.gov, MyVariant/ClinVar.\n- Materials Project.\n- FRED economic data.\n- Space weather.\n- Open-Meteo weather and climate.\n- USGS water data.\n\nYou can also add any local or remote MCP server from Settings. See\n[ docs/CONNECT_YOUR_TOOLS.md](/ai4s-research/open-science/blob/master/docs/CONNECT_YOUR_TOOLS.md).\n\nFor a neutral positioning note, see\n[ Open Science Desktop vs OpenScience](/ai4s-research/open-science/blob/master/docs/open-science-desktop-vs-openscience.md).\n\nDownload the latest installer from the\n[Releases page](https://github.com/ai4s-research/open-science/releases/latest).\n\n**macOS**:`.dmg`\n\n/`.app`\n\n, Apple Silicon and Intel, macOS 13 Ventura or later.**Windows**: NSIS`.exe`\n\nand`.msi`\n\n, Windows 10/11 x64.**Linux**:`.deb`\n\nand`.rpm`\n\non x86_64 Linux.\n\nBuilds are not code-signed or notarized yet.\n\n**macOS**: if Gatekeeper says the app is damaged or from an unidentified developer,\ninstall it into Applications and run:\n\n```\nxattr -cr \"/Applications/Open Science.app\"\n```\n\n**Windows**: if SmartScreen appears, choose **More info -> Run anyway**.\n\n**Linux**:\n\n```\nsudo apt install ./OpenScience_*.deb\n# or\nsudo rpm -i OpenScience_*.rpm\n```\n\nPrerequisites:\n\n- Node.js >= 20\n- pnpm 9\n- Rust toolchain\n- macOS, Windows, or Linux system dependencies required by Tauri\n\n```\ngit clone https://github.com/ai4s-research/open-science\ncd open-science\npnpm install\n\n# Fetch pinned sidecars and bundled skills. These are git-ignored.\nbash scripts/dev/fetch-opencode.sh\nbash scripts/dev/fetch-uv.sh\nbash scripts/dev/fetch-skills.sh\n\n# Run in development or build installers.\npnpm --filter @ai4s/desktop tauri dev\npnpm --filter @ai4s/desktop tauri build\n```\n\nUseful checks:\n\n```\npnpm test\npnpm typecheck\npnpm lint\n```\n\n- Workspace files, raw data, session history, provenance, notebooks, and run records stay local by default.\n- Command execution, file deletion, dependency installation, and remote connections are human-approved flows in the desktop app.\n- Provider credentials are written to app-private runtime config, not to the workspace, provenance, git, exports, or global OpenCode config.\n- Settings includes a plain-language data-flow view explaining what can be sent to the selected model provider.\n\n| Path | Purpose |\n|---|---|\n`apps/desktop/` |\nTauri + React desktop app. |\n`packages/sdk/` |\n`OpenCodeClient` ; keeps the UI from calling OpenCode directly. |\n`packages/shared/` |\nShared domain types and chart palette. |\n`packages/ui/` |\nShared UI package. |\n`runtime/skills/core/` |\nFirst-party scientific skills. |\n`runtime/skills/external/` |\nBuild-fetched external skills. |\n`runtime/harness/` |\nRuntime harness knowledge and operator context. |\n`runtime/mcp/` |\nMCP runtime notes/configuration. |\n`examples/` |\nBuilt-in example workspaces. |\n`scripts/dev/` |\nSidecar, `uv` , skill fetchers, and focused regression probes. |\n`docs/` |\nProduct, technical, operator, connector, and research notes. |\n\nThe project is a working desktop MVP in active development. The most reliable current\nimplementation log is [ PROGRESS.md](/ai4s-research/open-science/blob/master/PROGRESS.md). Product and architecture notes\nlive in\n\n[and](/ai4s-research/open-science/blob/master/docs/PRD.md)\n\n`docs/PRD.md`\n\n[, but those documents include target design as well as historical status notes.](/ai4s-research/open-science/blob/master/docs/TECHNICAL_DESIGN.md)\n\n`docs/TECHNICAL_DESIGN.md`\n\nNear-term work is focused on signed/notarized releases, broader Windows/Linux verification, auto-update, richer connector hardening, and continued reproducibility review.\n\nIssues and PRs are welcome. Keep changes minimal and verifiable, follow\n[ AGENTS.md](/ai4s-research/open-science/blob/master/AGENTS.md), and run the checks before opening a PR. For discussion,\njoin the\n\n[Open Science Discord](https://discord.gg/fWNMDKcd5P)or the\n\n[linux.do](https://linux.do)community.\n\n[MIT](/ai4s-research/open-science/blob/master/LICENSE). Bundled third-party skills and connectors keep their own licenses.\n\nOpen Science Desktop is beta research tooling. Treat outputs as drafts: verify numbers, citations, code, and conclusions before publication or decision-making.", "url": "https://wpnews.pro/news/show-hn-a-possible-open-source-desktop-alternative-to-openai-codex", "canonical_source": "https://github.com/ai4s-research/open-science", "published_at": "2026-07-10 10:30:48+00:00", "updated_at": "2026-07-10 11:05:46.184740+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-tools", "ai-agents", "ai-research"], "entities": ["Open Science Desktop", "OpenAI Codex", "Claude Science", "ResearchClawBench", "Tauri", "MCP", "OpenCode"], "alternates": {"html": "https://wpnews.pro/news/show-hn-a-possible-open-source-desktop-alternative-to-openai-codex", "markdown": "https://wpnews.pro/news/show-hn-a-possible-open-source-desktop-alternative-to-openai-codex.md", "text": "https://wpnews.pro/news/show-hn-a-possible-open-source-desktop-alternative-to-openai-codex.txt", "jsonld": "https://wpnews.pro/news/show-hn-a-possible-open-source-desktop-alternative-to-openai-codex.jsonld"}}