Huly Platform

Huly is an open-source platform built to eliminate the tool sprawl that burdens modern product teams. Instead of juggling separate apps for issue tracking, team chat, HR, recruitment, CRM, and documentation, Huly brings all of these into a single integrated workspace. Teams can manage sprints in the tracker, collaborate in real-time video rooms, write specs in the document editor, and hire candidates through the ATS — all without switching contexts.

The platform is architected as a distributed system of 30+ microservices, each responsible for a specific domain: a transactor service processes all data mutations with WebSocket-based real-time delivery, a collaborator service handles concurrent document editing using Y.js CRDT, and dedicated services manage full-text search, media processing, PDF generation, digital signatures, and event streaming. This separation means each domain can scale independently and be swapped without disrupting the rest of the platform.

Huly is built on CockroachDB for distributed SQL storage, Elasticsearch for search, MinIO for object storage, and Redpanda (Kafka-compatible) for async event streaming. The frontend is a 2,100+ component Svelte application managed in a Rush monorepo with 192 plugins, giving teams the flexibility to enable only the modules their organization needs.

Self-hosting is supported via Docker Compose through a companion repository (huly-selfhost), and a managed cloud version is available at huly.io. The platform also exposes a typed API client for building custom integrations and automations on top of the Huly data model.

What You Get

• Issue Tracker with GitHub Sync - Create and manage issues with priority, assignees, milestones, and status workflows that bidirectionally sync with GitHub Issues and Projects in real time.
• Virtual Office with LiveKit Video - Spatial office floors where team members appear as avatars in rooms with full audio/video conferencing, screen sharing, meeting recording, and live transcription in 40+ languages.
• Y.js Collaborative Document Editor - Real-time collaborative editor with CRDT-backed conflict resolution, inline mentions, code blocks, table of contents, PDF export, and action item assignment.
• HR and Recruitment Modules - Built-in HRM for org charts, departments, vacation tracking, and time-off management, plus an ATS with resume parsing (supports HH, LinkedIn, and generic formats) and candidate pipelines.
• Chunter Team Chat - Slack-like channels and direct messages with thread support, reactions, file attachments, and deep links into tracker issues and documents.
• Process Automation Engine - Define multi-step automation workflows with conditional logic, field locking, approval requests, and sub-process cancellation to automate repetitive operations across any module.
• Full-Text Search with Rekoni - Elasticsearch-backed search with a dedicated document intelligence service that extracts structured text from PDFs, DOCXs, and RTF files for searchable content across the workspace.
• CRM and QMS Modules - Customer relationship management with contact tracking and a Quality Management System with controlled documents, version approval flows, and audit trails.

Common Use Cases

• Running a distributed product team - A product manager creates sprint milestones in the tracker, links them to GitHub repos for automatic issue sync, holds standups in the Virtual Office, and stores PRDs in Huly Documents — all in one tab.
• Managing end-to-end recruitment - An HR director uses the ATS to post roles, pipeline candidates through stages, parse uploaded CVs with Rekoni, and onboard accepted hires into the HRM module without leaving the platform.
• Operating a regulated QMS - A quality team uses controlled documents with sequential review and approval workflows to maintain ISO-compliant documentation, versioning, and audit history entirely within self-hosted Huly.
• Building a custom business app - A developer uses Huly’s typed API client and plugin system to build a custom CRM workflow on top of the platform’s data model, leveraging existing auth, real-time delivery, and storage infrastructure.
• Coordinating a remote engineering org - An engineering lead connects multiple GitHub organizations to Huly, auto-creates tracker issues from GitHub Issues, assigns work in the team planner, and reviews progress in the centralized inbox.

Under The Hood

Architecture Huly is structured as a distributed monorepo containing 30+ independently deployable microservices organized into functional layers — core transaction processing, storage, real-time communication, media, search, and feature services. The central transactor service acts as the primary mutation engine, maintaining WebSocket connections with all clients and publishing domain events to a Redpanda (Kafka-compatible) message bus consumed by downstream services like the fulltext indexer, process engine, and media processor. Real-time document collaboration is isolated in a dedicated collaborator service using Y.js CRDT, ensuring that concurrent edits are conflict-free without burdening the main transaction path. Frontend logic is encapsulated in 192 plugins — each plugin owning its own models, UI components, server handlers, and assets — which allows modular feature enablement and clean domain boundaries at scale.

Tech Stack The platform is written primarily in TypeScript with Svelte 4 for the frontend, managed in a Rush monorepo with pnpm and built via Webpack 5. The primary database is CockroachDB providing distributed ACID SQL storage; Elasticsearch handles full-text indexing; MinIO provides S3-compatible blob storage for attachments and recordings; Redis backs real-time pub/sub for HulyPulse WebSocket notifications; and Redpanda delivers async event streaming between services. The virtual office uses LiveKit for WebRTC-based video and audio with HLS transcoding via a dedicated stream service. Backend services run as Node.js processes in Docker containers, with Jaeger providing distributed tracing across all service boundaries. The desktop client is packaged with Electron.

Code Quality The repository contains extensive automated testing across multiple layers: over 280 unit and integration test files using Jest alongside more than 100 Playwright end-to-end specification files covering HR, contacts, login, workbench, and tracker workflows. TypeScript is used in strict mode with interfaces over types, explicit error subclasses, and JSDoc annotations on public APIs. The Copilot instructions enforce consistent conventions — kebab-case files, PascalCase components, UPPER_SNAKE_CASE constants, no untyped any, no console.log in production — with automated commit validation and a changelog driven by git-cliff. CI pipelines run tests and svelte-check validation before merges to staging and main branches.

What Makes It Unique Huly’s most distinctive technical choice is its metadata-driven plugin architecture: each of the 192 plugins declares its own data models, UI presenters, and server-side handlers through a shared platform registry, allowing the frontend to auto-discover and inject components at runtime without hardcoded imports. The virtual office module goes beyond typical video conferencing by introducing spatial floor layouts where user presence, room states, and recording status are all synchronized through the same transactor event pipeline as issue updates and document edits. The rekoni service provides on-platform resume intelligence — parsing PDFs, DOCXs, and LinkedIn exports into structured candidate data — eliminating the need for external HR AI integrations. These capabilities combine to make Huly a genuine application platform rather than a bundle of loosely coupled tools.

Self-Hosting

Huly is released under the Eclipse Public License 2.0 (EPL-2.0). The EPL is a weak copyleft license: you can use, modify, and distribute Huly commercially, but if you distribute modified versions of EPL-licensed files, you must release those changes under EPL-2.0 as well. Crucially, the copyleft obligation applies only to the EPL-covered files themselves, not to code you write that merely uses or calls Huly — so building proprietary integrations or running Huly as internal infrastructure does not trigger any source-disclosure requirement. The license is business-friendly for self-hosters who want to run the platform without contributing back, as long as they do not redistribute modified EPL source.

Running Huly yourself is a meaningful infrastructure commitment. A full deployment involves 30+ Docker containers — transactor, workspace service, collaborator, fulltext indexer, rekoni, datalake, hulypulse, process engine, backup, print, sign, media, stream, and more — on top of CockroachDB, Elasticsearch, MinIO, Redpanda, and Redis. The README recommends Linux or macOS with Docker Compose and notes that a complete local build requires over 35 GB of WSL virtual disk space and a 4.5 GB source tree. There is a minified deployment profile that excludes non-essential services, but even that involves considerable orchestration. Teams choosing self-hosting are responsible for database backups, Elasticsearch index management, MinIO replication, certificate management for the sign service, and rolling upgrades across a versioned microservice graph.

Huly offers a managed cloud version at huly.io with tiered billing plans (Common, Rare, Epic, Legendary) that differ by storage and traffic quotas. The cloud tier handles infrastructure, upgrades, backups, HA, and SLA management — things a self-hosted deployment must address manually. The self-hosted path is well-supported through a dedicated companion repository (huly-selfhost) with Docker Compose configurations and an active GitHub community, but there is no commercially backed enterprise support tier or SLA for self-hosted deployments. Teams evaluating self-hosting should weigh their operational capacity against the breadth of the service graph before committing.