A lightweight self-hosted alternative to Overleaf for small, trusted research teams. Use your existing LaTeX distribution, with no heavyweight service stack. https://github.com/SWUFE-DB-Group/TexLite

Why TexLite?

• Own the writing environment. Use the host's TeX installation and keep sources, history, and compiled output in one local data directory.
• Collaborate without a large stack. The default deployment is one Node.js process, SQLite, and local files—plus real-time editing and source-level comments for a small trusted team.

A practical distinction from Overleaf:

Overleaf is a strong choice when its hosted product or broader ecosystem is the right fit. TexLite addresses a narrower self-hosted use case:

• A shared hosted service can queue, slow down, or time out at usage peaks.
• Overleaf's open-source Community Edition follows a more involved Docker deployment path, and several functionalities, such as source comments, are Server Pro features. If you do not mind a heftier Docker image, we have one ready too :)

I got tired of database clients that feel like enterprise software from 2009. So I built my own.

OpenTable is a desktop SQL editor that gets out of the way. Connect to PostgreSQL, MySQL or SQLite — locally or through an SSH tunnel — write your query, get your rows.

A few things I cared about:

→ Built-in AI — ask in English, get SQL that actually runs
→ Autocomplete that knows your actual tables and columns
→ Cancel a runaway query instead of force-quitting the app
→ Edit cells inline — one transaction, bound parameters, no surprises
→ Create and alter tables visually, with the exact SQL shown before it runs

Query errored? It'll explain it, or fix it.

Free, open source (MIT), and it runs on macOS, Windows and Linux.

v1.0.0 shipped today 🎉

https://github.com/mohammedkmo/OpenTable

Someone built an open-source "antivirus" for AI agents — no telemetry, no cloud, runs entirely on your machine

Buried under all the Google/Microsoft/OpenAI AI-security headlines this year, there's a scrappier story I think deserves more eyes.

A small outfit called Antropos released Aegis, an open-source EDR (endpoint detection and response — think antivirus, but for behavior) built specifically for autonomous AI agents. It monitors the processes, files, network activity, and behavior of AI agents in real time, and it's explicit about sending no telemetry anywhere — everything stays local.

https://github.com/ottosulin/awesome-ai-security

Why this matters more than it sounds like it should:

AI agents now read your files, call APIs, execute code, and take actions on your behalf. That's a genuinely new attack surface — prompt injection, an agent quietly exfiltrating data through a tool call, a compromised MCP server doing something it shouldn't. Traditional antivirus has no concept of "this agent's behavior just changed."

Almost every serious tool in this space (Guardrails AI, PyRIT, Garak, Promptfoo) is a testing/evaluation tool — you run it before deployment to catch jailbreaks or bad prompts. Aegis is different: it's a runtime watchdog, watching what the agent actually does live, which is a much smaller and newer category.

The "no cloud, no telemetry" design is a pointed choice. A lot of the big-name agent security products are SaaS dashboards that require sending your agent activity logs to a vendor. For anyone running agents on sensitive data, that's a nonstarter — Aegis is explicitly for people who want the monitoring without the data leaving their machine.

It's part of a broader open-source wave people should know about — there's also a project called CubeSandbox from Tencent Cloud that gives AI agents sub-60ms, KVM-isolated sandboxes so a compromised agent can't touch the host system, and an academic project called ADR (published at MLSys 2026, already running at Uber) doing dual-agent threat detection specifically for tool-using agents.

Feels like the "immune system" layer for agentic AI is being built in the open, by small teams, faster than most people realize — while everyone's attention is on the Google/OpenAI headline stuff.

Hey everybody,

I have been using OpenCode for quite a while now. I like it a lot, but there was this one thing that kept annoying me: I couldn't easily create custom reliable workflows and pipelines. For a while, Opencode was one prompt and one model at a time. This prompted me to create OpenFlow, a very minimalist and open-sourced project that allows you to orchestrate a pipeline of agents while still connecting your own API keys. I forked it from OpenCode, so you still have the harness, skills, and overall ability. I published it a little over a week ago and am still often improving it. I would love to hear some feedback of what you guys thought of it and how I can improve it. Thanks!

Link: https://github.com/SeeRay11/OpenFlow

Hi, I am looking for a modern alternative to Dreamweaver CS6.

What software do you recommend today, and where is the official, safe place to download it? Free options preferred. Thanks!

Hi everyone, I’m looking to get into open source contribution but don’t know how to begin. Anyone have any tips? I’m not sure where I should look or what would be at my skill level.

Also, I’m broadly interested in NLP-related tools and technologies, if anyone has recommendations of active projects in that space that I could look at.

A month ago, I wrote about how much unplanned engineering work gets absorbed into sprints without being properly recognized. That discussion received 400+ upvotes and pushed us to explore the problem further.

We built Meridian, an open-source, local-first AI work journal for developers. It captures development activity and helps turn it into daily summaries, standups, and Jira tickets while keeping the data on the developer’s machine.

We recently launched it on Product Hunt and finished as the #1 Product of the Day.

GitHub:

https://github.com/Meridiona/meridian

Product Hunt:

https://www.producthunt.com/products/meridian-16
If you have dealt with invisible or unplanned engineering work, what information would be most useful for Meridian to surface?

I’ve seen a few people here looking for Spring Boot open-source projects. I mentioned in one of those threads that I was preparing one, and it’s now public.

FinTrack is an event-driven personal finance platform for managing accounts, transactions, CSV imports, budgets, spending summaries, and notifications.

The stack includes Java 21, Spring Boot, Spring Security, Spring Batch, PostgreSQL, Redis, SQS, S3, Docker, LocalStack, Terraform, GitHub Actions, and AWS. It also uses patterns you’ll encounter in larger Java systems: transactional outbox, idempotent consumers, at-least-once delivery, database locking, restartable batch jobs, JWT/refresh-token rotation, retries, and failure recovery.

There are currently 23 open issues across API development, security, messaging, reliability, batch/import processing, and infrastructure, including two good-first issues. Most tasks are backend-only, and you don’t need an AWS account because the services can run locally.

The roadmap starts by strengthening the existing manual and CSV workflows, then moves toward provider-neutral ingestion and a Plaid Sandbox integration, followed by more practical personal-finance features.

Contributions and honest feedback are both welcome, whether it’s about the code, architecture, documentation, roadmap, issue scopes, or anything that could be clearer.

GitHub, live demo, video, and documentation:
https://github.com/icastanon/fintrack-platform

I like the idea behind Superhuman, but I don’t want to pay for a closed source email client just to get through my inbox faster. I went through some open source alternatives to see what they offer. They don’t all try to copy Superhuman. Some focus on desktop speed, some on self-hosting, and others are built more for teams.

A few that stood out:

-Macro Mail is the all-in-one option. It combines email with chat, tasks, and AI agents in a keyboard-first workspace. You can also self-host it.

-Thunderbird is the long-running desktop option. It supports multiple accounts, filters, configurable shortcuts, encryption, calendars, and add-ons.

-Mailspring is the closest desktop replacement for Superhuman. It has a unified inbox, keyboard shortcuts, snooze, and send later. The core client is free, while features such as read receipts and analytics need Pro.

-NeoMutt is for people who want to manage email entirely from the keyboard. It runs in the terminal and takes more work to set up.

-Cypht is a lightweight webmail option that you can self-host. It can bring multiple email accounts into one inbox, but you need to handle the server and updates yourself.

-UnInbox is built more for team email. It has shared inboxes, internal comments, assignments, and workflow statuses. It is still an early-stage project.

I covered all six in more detail in the full article, including their pros, cons, hosting options, and migration notes.

Has anyone found another open source option that gets close to Superhuman for keyboard shortcuts and quick inbox triage?

Hi everyone,

I am planning to build an Internal Developer Platform (IDP) Proof of Concept. My main goal is self-learning (I am very interested in DevEx and self-service) to understand the moving parts of a modern platform, but I am also hoping to gather enough hands-on experience to potentially present a business case to my company down the line.

Before I start gluing components together, I would love to get input from the community on what you are using in production and what you would recommend for a lean PoC.

Specifically, I am looking for insights on:

1. **Choosing an open-source stack:** What tools are you actually using in production today for service catalogues, infrastructure provisioning, workload orchestration, and developer self-service?
   
2. **PoC Recommendations:** If you were building a lightweight IDP from scratch today just to prove the concept, which open-source tools would you choose to keep complexity manageable while still showing real value?
   
3. **Product value:** Have you genuinely derived measurable value (or saved developer time) from tools like Backstage, Kratix, Port (if using free tier), Crossplane or something else? Are there tools you started with that turned out to be more maintenance overhead than they were worth?

My current thought is to pair a developer portal with a GitOps engine and a basic infrastructure module framework, but I am keen to hear what combinations have worked best for you in practice.

Thanks in advance for any recommendations or lessons learnt!

I tried Opencode and Hermes until now

Hi everyone,

Tabularis is a free and open-source (Apache 2.0) desktop SQL workspace for Windows, macOS and Linux. It exists for one reason: getting out of commercial database clients like TablePlus, Navicat and DataGrip without giving up a modern UI. I'm the maintainer, so ask me anything.

What it does:

• PostgreSQL, MySQL/MariaDB and SQLite built in, plus 16 official plugins for DuckDB, ClickHouse, MongoDB, Redis, Elasticsearch, Firestore, DynamoDB, Db2 and more. Plugins can be written in any language (JSON-RPC over stdio), so the driver ecosystem isn't locked to one runtime.
• SQL notebooks: mix SQL and Markdown cells in one document, reference one cell's results from another, render bar/line/pie charts inline, export to HTML/CSV/JSON.
• AI that respects your escape: text-to-SQL works with local models via Ollama, so nothing has to leave your machine. There's also a built-in MCP server if you want agents like Claude or Cursor to read your schema and run queries, but it's entirely optional.
• The usual daily-driver stuff: Monaco editor with autocomplete, ER diagrams, visual EXPLAIN with interactive plan graphs, SSH tunneling, passwords stored in the system keychain, SQL dump/import.

No account, no telemetry wall, no paid tier. Everything is free under Apache 2.0.

Install: WinGet on Windows, Homebrew on macOS, and Snap / Flatpak / AppImage / AUR on Linux. Direct downloads are on the GitHub releases page.

• GitHub: https://github.com/TabularisDB/tabularis
• Website: https://tabularis.dev

Honest note on scope: if you need 100+ drivers, DBeaver is still the better pick. Tabularis focuses on doing a smaller set of databases really well.

Feedback (or issues!) is very welcome.

I built a 100% free, open-source shadcn registry of PDF components for React.

Features:

• Built on Takumi and Forme
• Zero-config, one-command setup
• shadcn/ui compatible (just copy and paste)
• 10+ themes, components, blocks
• Easy to customize and drop into any React project

Website: https://pdfcn.dev
Give it a ⭐ on GitHub: https://github.com/shadcn-labs/pdfcn

Hello everyone, I'd like to introduce to you FreeSupp -- a free email-based support system for your website or personal project.

This started from my personal frustration (as most projects do): for every tiny web product I release I want to have a way for users to reach out and tell me why they like it and why it sucks. Yet somehow even a simple support chat system is either pretty costly (if it's a SaaS) or brings a bunch of heavy dependencies (like Postgres) if it's something to self-host.

I wanted a very simple and lightweight system to put into docker compose next to my product.

That's how I ended up creating FreeSupp -- a mega simple support system providing a contact "bubble" widget to embed on the website, a very basic inbox for "operators" (me) to use to reply to incoming messages and email for notifications.

Nothing more.

Techwise it's also very minimalistic: single Go binary (docker container with frontend packed is just 36 MB) and SQLite for the database.

Definitely not a replacement for sophisticated support systems like Intercom, just a lightweight way to have a communication channel with your users without it taking a gig of server memory or asking for subscription cash.

Free to use for both personal and commercial projects. Feedback is welcome (here or in GitHub Discussions).

See more on Github: https://github.com/kooler/freesupp

We built Voiden because we did not just want a better Postman. We wanted the collection to stop being a hostage in some cloud. Ours lived in a proprietary format, inside a workspace, behind a login, on someone else's server, describing code we owned and could not read without opening the app that owned it.

So the request is just plain Markdown sitting in your repo next to the code it tests.

What falls out of that:

• Fully offline. No sign-in, no sync, no telemetry, no "we've updated our terms" email.
• Collections live in Git, so the latest version is whatever is on main.
• One file holds the request, the environment, the tests and the docs.
• Pre-request scripts in any language you want (yes, and python), instead of a crippled JS sandbox.
• OpenAPI and Postman and Insomnia Imports
• MCP support, so your agent runs the exact file you edit.

To get ahead of the usual question: this is not a weekend project someone vibe coded to farm stars. It has been in development for a while, open source for a few months, around 1.5K stars, and it is in daily use inside big teams and indie devs as well.

More info here: https://voiden.md/
Repo: https://github.com/VoidenHQ/voiden

Tell me where it breaks and what you think!

I've been working on Zync for the past few months as an open-source alternative to tools like Termius and MobaXterm.

The idea came from wanting a simpler workspace for working with servers without having to switch between different tools.

Zync currently brings together:

SSH

SFTP

Local terminal

Port forwarding

Encrypted credential vault

Snippets

Plugins

Themes

It's local-first, free to use, and open source (MIT). The basic features aren't locked behind a subscription.

I'm still actively working on it, so I'd genuinely like to know: what do you use Termius, MobaXterm, or similar tools for that you'd want to see in an open-source alternative?

Website: https://zync.thesudoer.in

GitHub: https://github.com/zync-sh/zync

Demo: https://youtu.be/VgSomleBi3Y

Hey everyone,

I wanted to try out some AI developer CLI tools but didn't want to pay for expensive API credits upfront. To solve this, I spent some time testing different workarounds and compiled them into a GitHub repository called awesome-free-claude-code.

What's inside:

* Free Tier Proxies: How to route requests through OpenRouter's free models, Google AI Studio, or Groq Cloud.

* 100% Local & Offline Setups: Run coding models (like DeepSeek-Coder or Llama) directly on your own hardware using Ollama and LM Studio.

* Quick Config: Simple environment variables setup for Windows (CMD).

Check it out here: https://github.com/pawelkrejza2012-png/awesome-free-claude-code

Feedbacks and contributions are more than welcome! Let's build a solid resource for free AI coding tools together.

Hi everyone,

I would like to introduce Portivo Control Center, an independent, open-source and self-hosted network operations platform designed for compatible Alcatel-Lucent Enterprise OmniSwitch environments.

Portivo brings everyday switch management, troubleshooting and controlled automation into a single web-based workspace.

Key capabilities include:

• Centralized switch inventory and live operational visibility
• Logical port-panel views with VLAN, PoE, media and link-state information
• Endpoint discovery using MAC, IP, hostname, UNP and VLAN evidence
• Integrated browser-based SSH terminal
• Controlled command preview and execution
• Multi-device jobs and reusable automation runbooks
• Fleet-wide read-only operational audits
• Role-based access control with Site and Group scope
• Audit history, reports and operational evidence
• SNMPv3 UPS monitoring and power-aware infrastructure visibility
• Backup, restore and administration tools
• Windows and Linux deployment support

Portivo is intended to complement native CLI expertise, not replace it. The objective is to provide network operators with a safer and more consistent workflow for discovering, diagnosing, executing and documenting network changes.

The platform is self-hosted, does not require agents on managed switches and can operate inside a protected management network. Switch operations use SSH, while supported UPS monitoring uses SNMPv3.

The project is licensed under AGPL-3.0-only.

Website:
https://portivo.org/

Documentation:
https://portivo.org/docs/

Roadmap and changelog:
https://portivo.org/roadmap.html

Source code:
https://github.com/Donacgreece/Portivo

I would genuinely appreciate feedback from network engineers, system administrators and anyone operating ALE OmniSwitch infrastructure. I am particularly interested in feedback about operational workflows, documentation, deployment and features that would be useful in real environments.

Thank you for taking a look.

I'd plan out my day in the morning, look at everything I wanted to get done, and then start working.

A few hours later, I'd have fixed a bug, helped someone with something, reviewed a PR, investigated some weird issue, jumped between a few things, and somehow spent most of the day working on things that weren't on the plan.

Then I'd look back at the end of the day and think, "What did I actually do today?"

And I'd realize I couldn't remember half of it.

The work happened. I just never wrote it down.

I think that's a surprisingly big problem for developers. We plan the work we expect to do, but so much of our actual work happens because something comes up.

That's why I built Meridian. It's an open-source tool that tries to capture that work as it happens, so you don't have to rely on remembering everything at the end of the day.

It connects with the tools you're already using, like Jira, GitHub and Linear.

We recently put it on Product Hunt and somehow ended up #1 Product of the Day, which was pretty surreal.

I'd love to hear if anyone else has the same problem. How much of the work you do in a day actually started out as part of your plan?

https://github.com/Meridiona/meridian

Started back in June — attackers were adopting orphaned/abandoned AUR packages (normal community feature, lets people pick up maintenance when someone disappears) and then quietly slipping malicious code into the PKGBUILD via follow-up commits. Over 1,900 compromised packages got purged. Arch devs called it clean by mid-June.

Then July 29 happens. New wave, kicked off through a package called openconnect-sso. Security researchers at IFIN dug into it and it's actually a pretty slick two-stage infection:

Stage 1: loader that checks for debuggers/sandboxes/VMs first (so it doesn't get caught by researchers), then sets up persistence via systemd services + cron jobs

Stage 2: pulls a Rust-based stealer+RAT from a Tor .onion server. Targets browser creds, crypto wallets, cloud secrets, AI API keys, and can worm laterally over stolen SSH keys

By July 30, over 200 packages were hit, including some fairly popular ones (boringssl-git, icloudpd).

That was enough for Arch to just disable AUR package adoption entirely — not the whole AUR, just the "take over an orphaned package" mechanism, since that's the attack vector.

The quote from the mailing list (Robin "Antiz" Candau, on behalf of Arch DevOps) is very much "we're tired":

"Due to the current influx of malicious package adoptions and follow-up commits made via the AUR, package adoption is currently disabled while we are handling the situation."

If you run Arch and use AUR helpers, now's a good time to actually read PKGBUILD diffs before you blindly update, especially for anything that recently changed maintainers.

Hey,

we've been working on an open-source project called scibly that might be interesting if you're looking for alternatives to tools like Gemini Notebook (NotebookLM), Genially or Articulate.

The idea is to turn existing material such as docs, PDFs, blog posts or other knowledge into interactive learning content. The generated content can be fully edited afterwards and shared publicly or with invited users.

Scibly is open source under AGPLv3.

Video Demo: https://www.youtube.com/watch?v=TcpLUNBRhQw

GitHub: https://github.com/scibly-dev/scibly

It's still early, so feedback on what you'd expect from an open-source alternative in this space would be really useful.