Workspace
Blog

Beeblio Is Open Source

30 Oct 2026·

Meet Beeblio’s Apache 2.0 local edition: how it works with your files, how it differs from the hosted workspace, and how to try or contribute.

Beeblio started with a simple frustration: research work is spread across documents, references, datasets, scripts, and chats, while the tools meant to help often see only one piece at a time. The Beeblio OSS brings those pieces into one workspace on your own computer. Today, we're opening its code under the Apache License 2.0 and inviting researchers and developers to use it, inspect it, and improve it with us.

This post explains what the open-source Beeblio is, how it relates to Beeblio on the cloud, and where to begin if you'd like to try or contribute to it.

A research workspace built around your files

You begin by linking an existing project folder. Beeblio works with the files in that folder rather than moving your research into a new proprietary format. You can still open and edit those files with other tools. Beeblio adds a workspace around them: a file explorer, literature search and library, artifact views, an editor, and a project conversation with an AI research agent.

The parts are designed to work together:

  • Write with sources in view. The document editor supports Markdown, tables, math, images, inline citations, and a generated bibliography. You can search for literature or insert a citation while writing, then export to formats including DOCX, LaTeX, and portable Markdown.
  • Ask an agent to work in the same project. Chat can use the open file, selected text, mentioned files, and uploads as context. The Eve agent can inspect and edit project files, search literature and the web, update bibliographies and literature matrices, and run local Bash or Python workflows.
  • Keep the work together. References, data, analyses, reports, figures, and conversations remain connected to the project folder. A saved conversation is there when you return to the project.

For example, you might collect papers, add citations to a draft, ask the agent to check a bibliography entry, run a Python analysis on a dataset, and return to the resulting figure without switching to a separate research dashboard. The goal is a workspace that understands where your work lives and helps you move between its parts.

What “local” means here

The Next.js interface and Eve agent run on your computer. A linked project remains a normal folder on disk, and Beeblio keeps its own project and conversation data in local SQLite. The app is designed for one trusted user on one computer. Its local servers should stay bound to 127.0.0.1; the agent can execute commands with your user permissions, so you should treat its access to files as real access.

The AI agent needs an OpenRouter API key and a selected model; model and some research features call external services (unless you provide your own LLM inference). Optional jobs, such as compiling LaTeX or converting Office files, depend on tools installed on your computer. The README lists current requirements and setup details.

How it relates to Beeblio Cloud

There are now two ways to use Beeblio, with different operating models:

Open Source EditionHosted Beeblio Cloud
Where it runsOn your computerOn Beeblio's hosted infrastructure
Project filesLinked folders on your computerHosted workspace storage
App dataLocal SQLiteHosted database
Intended setupOne trusted userWeb service with accounts

The local open-source edition is a good starting point if you want to work directly with folders on your computer, inspect the code, or develop your own research workflow. The hosted workspace is there for people who prefer a cloud service without running the app locally.

Why open-sourcing the local edition?

Research workflows are personal. One project may live in Markdown and Python; another may depend on Word, BibTeX, and a particular folder structure. Publishing the local edition lets people see how Beeblio handles those workflows, adapt it to their needs, and propose improvements that others can use too.

The repository is licensed under Apache 2.0. The source license covers the code; it does not grant the Beeblio name or logos as branding for another product. You can read the license and repository notes before building on it.

Try it, then tell us what gets in your way

If you want to use the app, start with the GitHub Releases page. The desktop edition targets Apple Silicon Macs and Windows PCs; check the latest release's assets for installers. The desktop app does not require a Node.js checkout. You will still need an OpenRouter key and model for the agent.

If you want to develop it, use Node.js 24 and pnpm 11, then follow the run locally guide:

pnpm install
pnpm dev

Open http://127.0.0.1:3000, configure the model in Settings, and link a project folder. A small project you know well is the best way to find where the workflow helps and where it still needs work.

Ways to contribute

You do not need to start with a major feature. A bug report with clear reproduction steps, a documentation correction, a focused fix, or feedback from a real research workflow can be valuable. If you have a larger idea, open an issue before implementing it so we can discuss the approach.

The contribution guide covers local checks, pull requests, and contribution terms. Please use the private reporting path in SECURITY.md for vulnerabilities instead of posting details in a public issue.

We are at the beginning of this open source effort. If Beeblio could make one part of your research workflow less fragmented, we'd like to hear which part. Explore the repository, try it with your files, and help decide what we build next.

Beeblio Team

open-sourceresearch-workflowbeeblio

The Premise

Why we're building a full-stack research workspace — and finishing the job the IDE started.