Transcription
Today we're diving deep into the top trending open-source GitHub projects. This week, part two, we'll explore agents.m for making your code AI friendly, Pluto printint for lightning fast HTML to PDF conversion, and Parliament for building AI agents that truly follow your rules. Get ready to discover Buddybase for rapid internal app development, Moia for unified back-end systems, and fast API MCP to transform your APIs into AI tools. These projects are set to boost your productivity and innovation.
Welcome back to Manu AGI tutorials. Here we explore the exciting world of AI, latest AI tools for you. So don't forget to hit that subscribe button and the notification bell so you don't miss out on the latest AI insights. So let's start today's video.
Project number one, agents.m. A machine friendly README that lets AI developers understand your code. Agents.m is a simple, lightweight markdown file designed specifically to guide AI-powered coding assistants and agents in working smoothly with your project. It's like a README, but not for human readers. Instead, it's a clear, predictable place where you explain the structure of your project, how to set things up, how to test, and what style guidelines to follow. This dedicated file ensures AI agents like OpenAI's Codeex or other tools can dive right into contributing code with minimal confusion.
What makes agents.m unique is its laser-focused purpose. Often README files are written for developers and may not include the instructions an AI actually needs: how to run tests, what commands to use, or how to adhere to linting rules. agents.m fills that gap by providing concise instructions tailored for machine parsing. For example, commands to install dependencies, how to run tests, linting rules, PR title formats, or branch-specific workflows. And it's not one-size-fits-all. It supports monorepos too, letting you place specialized files in subdirectories so agents always pick up the most relevant guidance.
Beyond just helping AI understand your codebase, agents.m brings consistency and reliability by standardizing how agents are informed across Codeex, Factory, Gemini, AMP, and more. It smooths out tool differences and enables predictable agent behavior. Before the emergence of this format, each AI tool relied on loosely structured READMEs, often leading to inconsistent results. Thanks to agents.m, agents can follow clear instructions, run checks, and respect project conventions with far greater accuracy.
Agents.m also supports safety and quality. You can include instructions for running test suites, executing linters, or validating changes before creating pull requests. These safety rails encourage agents to self-verify their output, reducing bugs and promoting trust in AI-generated contributions. In short, agents.m turns AI from a guessing robot into a well-informed teammate. It knows how your project works, how to contribute correctly, and how to keep your codebase clean. Drop this file into your repo, and let your coding agents follow your rules, not lose their way.
Project number two, Pluto printint. Lightweight, high-quality conversion from HTML to PDFs and images. Pluto printint is a sleek and easy-to-use Python library that turns your HTML or XML content into beautifully rendered PDFs or crisp images without the bulk or complexity of full browser engines. Powered by the lightweight Pluto book rendering core, it delivers fast, high-quality output without needing Chromium, WebKit, or any browser dependencies.
What makes it unique is its simplicity and portability. Normal browser-based tools are often heavy, slow to set up, and bloated. Pluto printint rejects that entire approach. Instead, it uses a compact C++ engine wrapped in Python, so you get precision rendering that's quick to install and easy to integrate into scripts, batch processes, or automation pipelines.
Another key advantage is its focus on paged media. Pluto Book understands layout rules like page breaks, margins, and document structuring systems, resulting in print-quality formatting that handles things like invoices, tickets, reports, or even chart snapshots with style and accuracy. Whether you need a multi-page report or a single-page image, Pluto Print delivers content that looks polished and professional.
It also has excellent integration with data visualizations. For instance, if you create charts using libraries like Matplotlib, Pluto Print can render those charts directly into documents, seamlessly blending dynamic visuals with styled content. Users appreciate how lightweight and predictable the tool is. One user contrasted it with headless browser alternatives and highlighted that Pluto print is much lighter, no browser dependency, just a compact C++ engine with a Python wrapper delivering fast, portable performance without the overhead.
In essence, Pluto Print is distinguished by its combination of speed, elegance, and reliability. You get professional-looking PDFs and images straight from HTML or XML with minimal setup and maximum control. Perfect for creators, developers, or anyone looking to streamline content generation.
Project number three, Parlant. Reliable rule-driven AI agents that actually follow instructions. Parlant stands out by giving you control over AI behavior in real customer-facing interactions. Rather than crossing your fingers and hoping your AI follows a system prompt, Parlant lets you define clear, natural language rules called guidelines that the agent must obey. Each time a user interacts, the agent checks these guidelines to respond exactly as intended, yielding more consistent and trustworthy behavior.
Another standout feature is the concept of journeys. Structured conversational flows that guide users step by step. You set the path, the agent follows it, adapting to context along the way. It's not just a rigid script. It's flexible, context-aware, and dynamic. This means your AI can handle complex, multi-stage interactions like booking appointments or resolving support requests with both precision and nuance.
And here's where Parlant really shines: its enterprise-grade readiness. The latest version, Parlant 3.0, packs major enhancements, including dramatically improved response speed, preamble messages to make interactions feel instantaneous, and hardened APIs designed for the real world. Whether you're in finance, healthcare, legal, or any domain where accuracy matters, Parlant helps you scale AI safely and reliably.
Beyond behavior control and performance, Parlant is also explainable. Every decision, why a certain guideline triggered, which part of the journey the agent is at, it is transparent and traceable. If you need to audit or refine behavior, it's all there. In short, this tool is unique because it lets you build AI agents that behave exactly as you design, even under real-world complexity. If you want agents that are consistent, compliant, fast, and explainable, Parlant delivers.
Project number four, BuddyBase. Internal tools made fast, flexible, and free. BuddyBase is unique because it combines openness, speed, and flexibility into a single platform that radically simplifies building internal apps and workflows. As an open-source low-code platform, it slashes development time, letting teams go from idea to app in just minutes. No heavy coding required.
You get to connect to a host of data sources such as PostgreSQL, MySQL, MongoDB, REST APIs, Airtable, Google Sheets, and more, or even start from scratch with its built-in database. That breadth of integration means whether you're tying into existing systems or building something brand new, BuddyBase adapts to you.
What sets BuddyBase even further apart is its polished interface and experience. It generates responsive single-page applications that perform smoothly across mobile, tablet, and desktop. You build using high-quality, pre-built UI components and templates. Everything looks good and feels professional right out of the gate.
Plus, it's not just about visuals. The platform includes powerful automation tools. You can trigger emails, fire webhooks, or create complex approval flows, all with just a few clicks, while still having the option to extend logic with scripts or plugins if you're more technical.
Security and scalability are deeply baked in. BuddyBase is ISO 27001 certified and built for enterprise environments. It supports advanced capabilities like SSO, role-based access control, backups, audit logs, and can be hosted either in the cloud or self-hosted using Docker, Kubernetes, or Digital Ocean. Perfect for teams that need full control over data and infrastructure.
Another standout quality is its truly open nature. BuddyBase is released under GPL and MPL licenses, meaning you can modify, extend, or self-host it as needed. Developers can contribute, build plugins, or customize components freely, making it ideal for organizations that want a collaborative, evolving tool instead of a closed, rigid product.
In summary, BuddyBase stands out by offering speed, design quality, extensibility, and enterprise-grade control, all wrapped in an open-source, approachable platform. It empowers teams to build anything from approval forms to full admin dashboards quickly, securely, and exactly how they need it.
Project number five, Moia. A unified event-driven back-end framework that brings everything together. When complexity strikes your backend, it's usually because you're juggling separate frameworks: one for APIs, another for background tasks, yet more for event queues or AI integrations. Moia flips that script. It integrates APIs, background jobs, scheduled workflows, real-time streams, and even AI agents into one coherent system with shared state, observability, and workflow logic.
The brilliance lies in its core building block, the step. Think of every backend component, API endpoint, cron job, AI agent as a step. Each step can be triggered in different ways, like an HTTP call, a timer, or an event, seamlessly chained into powerful event-driven flows. It's like backend development remodeled to be componentized, logical, and elegant.
Handling different programming languages is often chaotic in complex systems. But Moia simplifies that too. Want Python for your AI logic and TypeScript for your HTTP layer? Both live harmoniously in the same workflow, sharing state and observability. No language silos, just functional flexibility.
Also baked into Moia is a full observability experience: structured logs, tracing, live state, visual debugging via the workbench. All available locally and in the cloud. Know exactly what runs when, how data flows, and where things break without stitching your own tooling. It's designed to scale. Steps are independently scalable, allowing parallel execution without monolithic bottlenecks. Fault-tolerant, retry-capable, and resilient by default. No extra infrastructure required.
The comparison to React is telling. Just as React made UI into composable components, Moia turns backend logic into composable, observable, scalable steps. If you're tired of backend fragmentation, Moia offers a unified engine, simplifying, streamlining, and empowering you to build backend systems that are powerful yet clean.
Project number six, Fast API MCP. Seamlessly transform your Fast API endpoints into AI-ready tools. Fast API MCP shines because it bridges your existing Fast API app with AI agents, turning endpoints into interactive tool-like resources. This isn't just an OpenAPI to MCP converter. It's a native Fast API extension, allowing your app's endpoints to double as Model Context Protocol (MCP) tools with virtually no setup. You simply point it at your Fast API application, and it auto-discovers your endpoints, preserves schemas and documentation, and mounts an MCP server, all without rewriting your API or adding layers of complexity.
One of its strongest features is built-in authentication via Fast API dependencies. That means you can secure your MCP tools using the same mechanisms you already trust and use. No custom security modules required.
Another standout is the minimal configuration. You don't need to configure OpenAPI or MCP settings manually. Just attach Fast API MCP to your app, mount it, and the MCP interface spins up, complete with the full Swagger documentation intact and ready for AI agents to consume.
Deployment flexibility is also a highlight. You can mount the MCP server within the same Fast API application or deploy it separately if you prefer isolation or a multi-service architecture. Communication is efficient, too. Fast API MCP uses the ASGI interface directly, not external HTTP calls, making internal transport smooth and performant.
Beyond the technical conveniences, its synergy with modern AI tooling is compelling. Fast API MCP turns your web API into something AI agents can discover and call as tools. Many AI frameworks and agents, like those using Langchain or Claude, use the MCP standard to locate and invoke functions. Fast API MCP makes your endpoints instantly accessible to those agents without requiring scaffolded adapters.
Altogether, it's the union of zero-config deployment, secure integration, flexibility in architecture, and seamless compatibility with AI ecosystems that makes Fast API MCP a truly unique and powerful way to expose your API logic directly to intelligent agents.
Project number seven, Magic. The first open-source all-in-one AI productivity platform. Magic isn't just another AI tool. It's a full ecosystem. What sets it apart is that it's truly comprehensive, offering everything a business needs in one open-source package: a general-purpose AI agent, Super Magic; a visual workflow builder, Magic Flow; AI-driven instant messaging, Magic IM; and soon a collaborative office suite, Team Share OS.
At its heart, Super Magic is an AI agent built for complex, real-world tasks. It understands your instructions, plans what to do, takes action, and even corrects its own mistakes, all independently. It's like having a smart assistant you can tailor to your specific business needs.
Then there's Magic Flow, an intuitive drag-and-drop visual builder where you can design AI workflows. It supports integrations with various messaging platforms like WeCom, DingTalk, and Feishu, and works seamlessly with the other Magic components. It also offers powerful debugging and real-time monitoring to keep your AI apps reliable and responsive.
Magic IM brings intelligence into team communication, acting like AI-powered messaging that understands context, retrieves from enterprise knowledge bases, manages conversations, and supports multi-person group chats while keeping data isolated and secure per organization.
What truly makes Magic unique is how it integrates these pieces into a unified, open-source productivity platform. It not only delivers each component on its own terms but also provides the building blocks like AgentLang (build AI agents through natural language) and Magic Lens (convert complex HTML to clean markdown) so developers can extend and compose new workflows and tools.
In short, Magic stands out because it's not just a single AI tool. It's a cohesive, modular platform that lets organizations build, customize, and deploy AI-powered systems across workflows, communication, and collaboration, all within a single open-source environment that scales with your vision.
Project number eight, Frappe HRMS. A completely open-source, full-feature HR and payroll suite built for growth. Frappe HRMS stands out because it's one of the few truly open-source HR and payroll systems designed from the ground up to be flexible, affordable, and powerful. It combines more than a dozen key HR modules from employee onboarding, attendance, leave to payroll, taxation, performance, and more, all in one clean platform. No hidden fees, no vendor lock-in. You get total control over your system, all supported by a vibrant developer community and hundreds of partners.
What makes it unique is how deeply customizable it is. Teams can tweak workflows, reports, forms, layouts, and print formats on the fly without needing any coding skills. Whether you're a small business or a growing enterprise, you won't face escalating user fees. The system scales with you effortlessly, saving you from overpaying as your team expands.
The modular design also means you can adopt what you need when you need it. Start with core tools like leave management or payroll. Then layer on performance tracking, recruitment, and KPIs later. Integration is another standout feature. Frappe HRMS integrates seamlessly with ERPNext accounting or other third-party tools like biometric systems, saving countless hours on reconciliation work.
It's also mobile-friendly. Managers and employees can use the mobile app to request or approve leaves, clock attendance with geolocation, view payslips, or access profiles on the go. Reports are rich and actionable. Track key metrics like headcount, attendance trends, appraisal outcomes, and payroll breakdowns, or build your own dashboard.
Finally, it's powered by the Frappe framework, delivering a modern user interface, REST APIs, and rapid application development capabilities, all backed by open-source transparency. In short, Frappe HRMS is unique because it gives teams full autonomy over their HR and payroll infrastructure. Completely open, deeply customizable, scalable, and integrated, all without breaking the bank. It streamlines operations so companies can focus on people, not paperwork, and grow without friction.
Project number nine, CAP. The open-source alternative to Loom. Record, edit, and share beautifully. CAP offers a fresh, open-source take on screen recording that puts power, privacy, and polish front and center. What makes it truly stand out is how seamlessly it combines the slickness of recording with flexible control over your content. It's lightweight, cross-platform, and privacy-focused. Whether you're on macOS, Windows, or using the web, it handles beautifully.
One standout feature is the choice between instant mode and studio mode. Instant mode lets you hit record and get a shareable link in seconds, complete with captions, titles, summaries, and chapters already generated. Studio mode prioritizes quality, letting you record locally, edit, and polish before sharing.
CAP doesn't lock you into its infrastructure. You can store your videos locally, use your own cloud storage, or connect a custom S3 bucket, giving you total control over where your recordings live. Overlays like reactive comments, viewer analytics, and transcripts let you turn recordings into real conversations, virtually replacing quick sync meetings.
Its open-source nature isn't just a label, it's a philosophy. You can inspect the code, contribute new features, or even self-host via Docker. Built using a modern tech stack, including Rust, TypeScript, Next.js, Tauri, Drizzle, and Tailwind CSS, CAP reflects a clean, modular architecture that's both efficient and extendable.
CAP also delivers high production value. Think 4K resolution, 60fps capture, and smart compression that balances visual fidelity with file size. In a world where most tools lock your content into someone else's cloud, CAP gives you elegance and ownership. It brings professional features, beautiful presentation, and data sovereignty together in a single package. For anyone seeking a powerful, privacy-respecting way to record and share screen content, CAP is a standout solution.
Project number 10, Modern Monaco. A modern, effortless embeddible code editor with built-in language support. What makes Modern Monaco unique is how it transforms embedding a powerful editor into a truly seamless experience. No complicated setup, just built-in smarts out of the box. It delivers full Language Server Protocol (LSP) support for common languages like HTML, CSS, SCSS, Less, JavaScript, TypeScript, and JSON. So features like autocomplete, error checking, inline CSS, HTML, and JS, and even autoclosing tags work automatically without any extra wiring. No more fiddling around with worker configurations. The necessary LSP workers are autoloaded behind the scenes, freeing you from setup headaches.
Another standout is smart file behavior. It supports a file system provider that enhances import suggestions, and it understands embedded languages like HTML or CSS inside JavaScript or template files with context-aware help. So the editor isn't just stuck language by language. It understands how code lives in the real world.
Flexibility is a core strength. If you want to trim down your bundle size, Modern Monaco offers a core version via `modern-monaco/core` that lets you drop the built-in grammars and LSP providers. Perfect for leaner builds on lightweight projects. This thoughtful modularity means you're not forced to take unnecessary baggage.
All of this comes wrapped in a modern experience. An updated take on the Monaco editor that's easy to embed, smart by default, and friendly to performance-sensitive environments. No custom worker setup, no manual LSP integration, just slick editing out of the box with deep language awareness and a sensible modular path if you want to go minimal.
At its heart, Modern Monaco is unique because it bridges the gap between lightning-fast setup and powerful editor functionality. It strips away complexity while preserving the rich editing features developers love.
And that's a wrap on our second look at these amazing open-source projects. We've seen Magic, a comprehensive AI productivity platform; Frappe HRMS for flexible, full-feature HR and payroll management; and CAP, the privacy-focused Loom alternative for screen recording. Finally, Modern Monaco simplifies embedding powerful code editors. Each of these tools showcases the power of open-source, offering transparency, control, and cutting-edge solutions for developers everywhere.
If you found this video valuable, don't forget to hit that like button, subscribe for more deep dives, and tell us which project you're excited to try in the comments below.