Open Source App Cloud Marketplace

With over 330 GitHub stars and described as "OpenRouter, but for agent tools instead of models," Treg is the unified tool registry that gives your AI agent one base URL, one token, and instant access to approximately 2,800 cataloged endpoints across 56 providers — SEO and backlink analysis via Semrush and Moz, social media intelligence from TikTok and Instagram, people and company enrichment through Crunchbase and Hunter, media buying, scraping, and SERP data — all priced per call from fractions of a cent with no provider signup required. The FastAPI backend with PostgreSQL stores credential bindings and injects them server-side into every proxied request, ensuring API keys and OAuth tokens never touch the agent or local machine while every call is audit-logged. Teams register their own paid API accounts, OAuth connections, vendor CLIs like Stripe, GitHub CLI, Vercel, and gcloud, plus custom SKILL.md definitions — all callable by every teammate's agent through a single treg token with automatic credential injection. The evidence-based tool selector shows success rate, response speed, and last-seen timestamp for each provider, so agents pick tools on measured performance rather than guesswork. Bring-your-own-key routes bypass metering entirely, and the one-dollar free credit on every new team lets agents start calling immediately. The Vue.js web dashboard provides team management, audit logs, catalog browsing, and guided onboarding. Self-host with pip install tools-registry[server] on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. Apache-2.0 licensed.

Why keep a Java runtime on the server just to host Minecraft, when Pumpkin runs from a single native executable and lets Java and Bedrock players share a world? Java Edition clients join on port 25565 and Bedrock clients on 19132, although Bedrock support is still marked work in progress. Builds exist for Windows, Linux, macOS, and Android. On launch, the console reports the worlds loaded and the server ready within milliseconds. Plugins are sandboxed WebAssembly modules that raise console permission prompts and pass signature checks, and authors can write them in Rust, Kotlin, Python, Go, C#, C/C++, D, Zig, or TypeScript. The ppm package manager searches and installs community plugins from the Pumpkin Market, and PatchBukkit loads some Bukkit and Paper plugins, with expected gaps. One pumpkin.toml file holds ports, MOTD, view distance, whitelist, authentication, compression, PvP, chat, resource packs, and world options, while chunks save in Anvil, Linear, or the native Pump format. Velocity and BungeeCord forwarding fit proxy networks, and the Query protocol plus Minecraft Server Management Protocol cover monitoring and remote administration. A public parity tracker lists every vanilla block, entity, item, and command as implemented, partial, or planned, so gaps are easy to spot before a world moves. Today it suits private survival worlds, test servers, and plugin development. It is early software, so the project advises backups. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. GPL-3.0 licensed.

Viewers arrive on browsers, phones, and desktop players, and SRS meets each of them with the right protocol from a single published stream. A broadcaster pushing RTMP can reach browser viewers through WebRTC with sub-second latency, phone apps through HLS, and desktop players through HTTP-FLV, all from the same live source without running parallel encoders. WHIP endpoints accept WebRTC ingest from browser studios and tools like OBS, while WHEP serves WebRTC playback, so an entire publish-to-viewer path can stay on modern web protocols. Operators shape each channel with virtual hosts, authenticate publishers and players through HTTP hooks that consult your own backend, and record live streams to disk as DVR archives for replay. Forwarding republishes a stream to YouTube, Facebook, or another SRS instance, and origin clustering lets edge servers pull from a central origin so one ingest feeds viewers in several regions. The built-in HTTP API reports streams, clients, and server health for monitoring, and GB28181 support ingests surveillance camera networks alongside creator broadcasts. Configuration lives in a single reloadable file, so changing a vhost or enabling HLS rarely means restarting the listeners your audience is attached to. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.

Confluence and Notion both want your team's documentation in their cloud; Docmost, an open-source collaborative wiki platform, keeps it on your server. The centerpiece is a Notion-style block editor with CRDT-based real-time collaboration: multiple people edit the same page simultaneously and changes merge without conflicts or overwrites. Content lives in spaces - per team, project, or department - with nested page trees, role-based permissions, groups, inline comments, page history with restore, and full-text search across everything. The editor covers tables, code blocks, callouts, KaTeX math, and file attachments, and diagramming is built in rather than bolted on: Mermaid, Draw.io, and Excalidraw all render inside pages, alongside embeds for Airtable, Loom, Miro, and more. Migration paths include Notion, Markdown, HTML, and ZIP archive imports (Confluence, PDF, and DOCX importers ship in the Enterprise edition, along with SSO via SAML/OIDC/LDAP and MFA). The stack is TypeScript with PostgreSQL and Redis, deploys via Docker Compose, runs in air-gapped environments with no external dependencies, and is translated into 10+ languages. The AGPL-3.0 community edition carries no per-seat fees; the project has passed 20,000 GitHub stars since its 2024 launch.

Whether it is a delivery van, a trailer in a storage yard, or a dog on a hike, Traccar is the GPS tracking server that puts it on one live map you host and control. Every position and event lands in your own SQL database. Over 200 protocols covering 2000+ tracker models decode out of the box, so a Teltonika fleet unit and a budget dongle share one device list. Pick a device to see speed, address, ignition, and total distance. Routes replay on a timeline slider, and a ruler tool measures distances. Geofences can be polygons, polylines, or circles, linked to devices or groups, colored as you like, and limited by calendar schedules. Alerts for overspeed, fuel drops, offline devices, alarms, and maintenance arrive by web, email, push, SMS, Telegram, or WhatsApp. Reports cover trips, stops, summaries, charts, events, and positions, with Excel export and scheduled email delivery. Commands go out over the tracker's own connection or by SMS, queued for offline devices. Computed attributes derive ignition or fuel values from raw inputs using expressions. Resellers get admin, manager, user, and read-only roles with device limits, plus white-label branding. Traccar Client lets a smartphone report as a tracker, and Traccar Manager monitors devices from mobile. Teams who want AI assistants to query their devices can add the MCP endpoint, which is optional. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. Apache-2.0 licensed.

Dashboards should not wait minutes for yesterday's data, so Apache Doris answers analytical SQL over fresh data with a massively parallel, fully vectorized engine that processes column batches instead of row after row. Frontends receive queries over the MySQL protocol, plan them with a cost-based optimizer, and coordinate execution, while Backends store columnar tablets and run scans, joins, and aggregations in parallel pipelines across cores. Data lands continuously through stream and batch loads, broker imports from object storage, and change capture from operational databases, and it is queryable immediately for reporting that reflects the last minutes rather than the last nightly job. Tables outside Doris join the same SQL through catalogs for Hive, Iceberg, Paimon, Hudi, and JDBC sources, with local caching and materialized views accelerating the lakehouse queries your team runs every day. Search workloads live here too: an inverted index and a SEARCH function provide full-text retrieval, vector indexes support similarity search, and AI functions call language models from SQL for extraction and summarization, so hybrid queries mix keywords, semantics, and structured filters in one statement. Workload groups, resource controls, and compute-storage separation in cloud mode keep interactive dashboards fast while heavy ETL spills safely to disk. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. Apache-2.0 licensed.

Private npm packages without shipping code to the public registry or paying for npm Enterprise: Verdaccio is the standard lightweight, zero-config private registry and caching proxy. It runs as a single Node.js process with its own tiny embedded database; no external database is required to start. Point npm, yarn, or pnpm at it and everything behaves as expected: install, publish, unpublish, dist-tags, and deprecation all work against the standard npm endpoints. The uplink system is where it earns its keep: packages not found locally are fetched from configured upstreams (npmjs.org, yarn, JFrog, Nexus, or another Verdaccio), cached as tarballs, and served locally thereafter - cutting CI latency and surviving registry outages. Multiple uplinks chain for failover, and per-package glob patterns in config.yaml route scopes to specific upstreams while controlling access, publish, and unpublish rights per group. You can even override a public package by publishing a patched version under the same name locally. A plugin architecture swaps in auth backends (htpasswd default, LDAP and others available), storage drivers (S3, Google Cloud Storage), middleware routes, and metadata filters. With official Docker images and a Kubernetes Helm chart, it slots into any pipeline in minutes.

Estimators, planners, and site teams work from one project record in OpenConstructionERP, a self-hosted construction workspace that carries a job from priced bill of quantities through tender, schedule, and site diary. The BOQ editor organizes sections, positions, and reusable assemblies with overhead, profit, VAT, and contingency markups, validates against DIN 276, NRM, and MasterFormat, and exports to PDF, Excel, CSV, or GAEB XML. Pricing draws on 120,000+ cost items across nine cost bases and a 7,000+ resource catalog, with regional price adjustment and semantic search. Quantities come from PDF plans measured on screen or from RVT, IFC, DWG, and DGN models; a 3D viewer filters elements and links them to BOQ positions, clash detection checks federated models, and an AI estimator drafts positions from text, photos, or drawings via the LLM you connect. Gantt schedules show dependencies and critical path, while earned value and S-curves track budget against actual. Tender comparison lines up subcontractor bids, and contracts, variations, and interim payment claims stay tied to the contract sum. On site, daily diaries with health and safety tracking, punch lists, and a carbon module totaling embodied emissions from the BOQ keep field work in the same system. The interface speaks 44 languages with right-to-left support, 293 guided cases teach real workflows on a demo project, and a REST API, webhooks, and module builder extend it. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. AGPL-3.0 licensed.

To pull a team's skills, MCP servers, hooks, prompts, and agent files out of scattered Git repositories, Observal keeps them in one reviewed registry and traces real sessions to show which ones help. Build an agent visually, with a live config preview for every harness. Submissions wait in an admin review queue, where reviewers inspect side-by-side version diffs before approving a release. Teamspaces give each team its own namespace, and approved releases can be forked into independent drafts. One pull command then writes the right config for Claude Code, Cursor, Kiro, Codex, Copilot, Pi, OpenCode, Antigravity, or Goose. Every approved resource is discoverable through the open Agentic Resource Discovery standard, and a running agent can hand part of its task to another approved agent or a remote Agent2Agent service. Hooks record each session as a trace of prompts, tool calls, MCP round trips, tokens, and errors, with turn-by-turn replay for debugging and audits. Insight reports, generated through providers such as Anthropic, OpenAI, or Ollama, list what is working, what is hindering, and quick wins. A dashboard follows adoption, cost, and department usage, a leaderboard ranks agents by downloads, and audit logs, SAML single sign-on, and SCIM provisioning are included. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. Apache-2.0 licensed.

GPT Load has become the go-to self-hosted AI gateway for teams managing multiple LLM provider credentials behind a single URL. The Go-built proxy transparently preserves native API formats for OpenAI Chat Completions, OpenAI Responses, Anthropic Messages, and Google Gemini — applications swap one base URL and keep their existing SDK integration untouched. Twenty built-in channels cover official APIs (OpenAI, Anthropic, Gemini, xAI), cloud platforms (Azure OpenAI, AWS Bedrock, Google Vertex AI), model services (DeepSeek, Moonshot AI, SiliconFlow, Zhipu AI, Alibaba Cloud, Volcengine, OpenRouter, Groq), and subscription accounts (Codex, Claude, Antigravity, Grok) using the same credential management, scheduling, and health system. The intelligent key pool rotates among valid credentials using atomic counters for fair high-concurrency distribution, automatically blacklists failing keys after configurable thresholds, and recovers them via scheduled health checks — all transparent to clients. Weighted load balancing distributes traffic across multiple upstream endpoints while per-credential proxies route individual keys through different egress paths. The Vue 3 management dashboard provides real-time statistics, group configuration, key testing, request logs with full debugging context, per-model cost estimates, and route inspection. Docker deployment exposes the management interface and proxy on port 3001 with SQLite by default, MySQL or PostgreSQL optional, and Redis for distributed leader-follower cluster scaling. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.

Krayin CRM gives sales teams a visual Kanban pipeline where leads flow through configurable stages — new, contacted, qualified, proposal, won, lost — with deal values tracked per stage, drag-and-drop transitions, and AI-powered document import that extracts contact details from uploaded PDFs, DOCs, and images without manual data entry. Built on Laravel 12 with a Vue.js frontend and MySQL database, the platform delivers complete customer lifecycle management through a two-stage contact model separating unqualified Leads from qualified Persons linked to Organizations, with a unified timeline displaying associated activities, notes, emails, and deals. The activity module schedules calls, meetings, and tasks on a drag-and-drop calendar with reminders and team assignment. A built-in product catalog supports line-item quoting with pricing and quantities. Email integration connects via SMTP, IMAP, and SendGrid parsing, linked to leads and contacts automatically. Workflow automation triggers actions on conditions and pushes events to external systems via webhooks. The dashboard visualizes pipeline performance, top customers, top products, and email engagement with pie charts and trend widgets. Custom attributes add text, number, date, boolean, and dropdown fields to any entity. Role-based ACLs manage team permissions per module. Extensions add multi-tenant SaaS, WhatsApp lead generation, and VoIP calling. Deploy via Docker Compose. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.

Alertmanager forgets an alert the moment it resolves; Jarvis remembers everything. This web frontend watches all your Alertmanager clusters at once and streams firing alerts into one live view over WebSocket, no refresh as problems appear or clear. Every alert's lifecycle is stored, so the resolved view answers how often something fired and how long it burned, and a grace period keeps a missed poll from being recorded as a phantom resolution. When an alert lands, claim it so the team sees who is on it, and leave comments bound to that alert fingerprint; notes survive restarts and re-fires, so the next person inherits the context. Silencing gets a rehearsal: preview exactly which alerts a silence will hit before creating it, fire a one-click Fast-Silence from the list, and save templates for recurring maintenance. Filters use the Alertmanager matchers you already know, save as named sets, ride in shareable URLs, and one can be your default view. Point Jarvis at every member of an HA gossip cluster and duplicates collapse by fingerprint, while protected upstreams authenticate with OAuth2, bearer tokens or basic auth per cluster. Sign-in is optional through built-in accounts or any OIDC provider, and it runs as one hardened container on port 8080, with PostgreSQL available when you want replicas. Keep it behind your VPN or an authenticating proxy, as its own docs insist. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. Apache-2.0 licensed.

Why rent your messaging from a corporate walled garden when Tinode runs WhatsApp-style chats, groups, channels, and calls on a server you control? Teams get one-on-one and group chats that stay in sync across web, Android, and iOS, with typing indicators, delivery and read receipts, presence, replies, forwarding, editing, pinned messages, and notes to self. Communities broadcast through read-only channels with an unlimited number of subscribers, and granular permissions decide who can read, write, or invite on each topic. Support desks welcome anonymous visitors, mark staff and verified accounts with badges, and put chatbots to work with forms and templated replies. Messages take markdown-style formatting, inline images and videos, file attachments, and voice messages. One-to-one voice and video calls run peer to peer over WebRTC, using the TURN and STUN servers you configure. Administrators block unwanted contacts on the server, attach moderation plugins over gRPC, and script the rest with a command-line tool. Attachments can sit on local disk or Amazon S3, the database can be MySQL, PostgreSQL, or MongoDB, and sharded clustering with failover scales past one server. The clients offer light, dark, and system themes, custom chat wallpapers, and translations for more than a dozen languages. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. GPL-3.0 licensed.

DCM is a visual builder that eliminates docker-compose copy-paste anxiety for home server enthusiasts. The TypeScript and Next.js web interface on port 7576 presents a curated catalog of popular self-hosted applications organized into categories: media management with Jellyfin, Plex, Sonarr, Radarr, and Prowlarr; dashboards like Homarr and Heimdall; databases including PostgreSQL, MariaDB, MongoDB, and Redis; monitoring stacks with Grafana and Prometheus; download clients like qBittorrent and NZBGet; security tools including Vaultwarden; storage solutions like Nextcloud; and home automation via Home Assistant. Select any combination of containers and DCM generates both compose.yaml and .env files with correct volume mounts, environment variables, PUID/PGID ownership, network modes, and port mappings — ready to paste into your terminal. Automatic port conflict detection identifies clashing services and reassigns ports before you ever hit a deployment error. The Template Gallery ships predefined multi-container stacks for common setups: media servers, development environments, database clusters, and monitoring solutions — some incorporating ten or more services in a single template. Built as a static Next.js export served from a 62 MB multi-arch Docker image supporting amd64, arm64, and arm/v7, DCM runs on everything from rack servers to Raspberry Pis. Share generated configurations via URL or download them locally. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. AGPL-3.0 licensed.

GCHQ open-sourced its "Cyber Swiss Army Knife", and CyberChef became the web app security analysts, incident responders, and CTF players reach for when data needs decoding, decrypting, or dissecting. Its interface is four panes: paste or drag input (files up to 2GB), search a categorized library of hundreds of operations, drag them into a recipe with arguments, and read the output. Operations span Base64, hex, and XOR encoding; AES, DES, and Blowfish encryption; classical ciphers from Caesar to Railfence; hashes and checksums; compression; regex and string extraction of IPs, domains, and URLs; timestamp conversion; and parsers for IPv6, X.509 certificates, and more. Recipes chain arbitrarily - convert from a hexdump then decompress, decrypt AES pulling the IV from the cipher stream, or let the Magic operation auto-detect several layers of nested encoding. Auto Bake re-runs the recipe live as input or arguments change, Step executes one operation at a time for debugging, and flow control (forks, subsections, registers) applies different operations to different parts of the data. Recipes save to files or share as URLs encoding the full pipeline. Crucially, CyberChef is entirely client-side JavaScript - nothing uploads anywhere - and self-hosting guarantees an unmodified copy inside your own network, where malware artifacts belong.

GrowChief delivers a self-hostable social media automation engine that replaces expensive SaaS tools like PhantomBuster, Expandi, and Zopto with a Docker-deployable alternative you fully control. The React and NestJS monorepo architecture runs headful Chrome via xvfb with Playwright and Patchright for maximum stealth — natural mouse movements and randomized interaction patterns make automation indistinguishable from manual browsing. Temporal orchestration manages workflow execution with built-in concurrency control that throttles actions to one every ten minutes per account even when ten workflows trigger simultaneously, preventing rate-limiting and account flags. The enrichment waterfall resolves profile URLs from partial data — provide an email address or organization name and GrowChief queries multiple providers to locate the correct social profile automatically. Working hours scheduling ensures outreach runs only during configurable business hours regardless of when leads enter the pipeline via the public REST API. Proxy support accepts custom SOCKS5 and HTTP proxies per account for geographic distribution and IP isolation. The public API enables full workflow management and lead injection from n8n, Make, Zapier, or any HTTP client, with a dedicated n8n custom node shipping ready to install. The multi-step workflow builder lets you chain connection requests, profile visits, message sends, and conditional delays into repeatable sequences. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. AGPL-3.0 licensed.

Navigating organizational knowledge, visualizing concept graphs, and orchestrating cooperative AI agents converge inside Yuxi, an open-source multi-tenant workspace built for enterprise research and deep archive exploration. Users can upload diverse business documents including scanned PDFs, spreadsheets, and slide presentations to automatically extract structured text blocks, tables, and entity relationships for semantic querying. Knowledge teams can inspect interactive topological graph views to explore connections between company entities, cross-referencing subgraphs against original file sources. Practitioners can configure autonomous agents with specialized subagents, assigning dedicated roles for concurrent research, deep data synthesis, and code execution. Human reviewers can inspect agent thought chains in real time and evaluate interactive confirmation cards before agents execute irreversible external actions or modify shared workspace files. Operators can execute sandboxed tasks that produce downloadable reports, interactive data charts, and rendered HTML pages stored within isolated workspace file systems. Administrators can govern model endpoints, set granular permission boundaries across departments, and monitor token consumption through live system analytics. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.

With over 52,000 GitHub stars and its role as the foundation of every Kubernetes cluster worldwide, etcd is the distributed key-value store that handles the most critical data in modern infrastructure — cluster state, configuration, service discovery, and distributed coordination. Built in Go and graduated from the CNCF, etcd uses the Raft consensus algorithm to maintain a strongly-consistent, highly-available replicated log across a cluster of machines, gracefully handling leader elections during network partitions and tolerating machine failure including the leader node. The gRPC API provides atomic key-value operations including put, get, delete, and transactions with multi-key compare-and-swap semantics, enabling distributed locking, leader election, and configuration management without external coordination. Watch operations stream real-time change notifications for specific keys or key ranges, powering reactive architectures that respond immediately to configuration updates. Version 3.7 introduces RangeStream for streaming large result sets in chunks, keys-only range requests for faster metadata queries, and bootstraps entirely from v3store after eliminating the legacy v2 store dependency. Automatic TLS encryption with optional client certificate authentication secures all cluster communication, while role-based access control restricts key access per user. The embedded bbolt B+ tree storage engine provides consistent reads and writes with configurable compaction policies. Benchmarked at 10,000 writes per second per instance with linearizable reads, etcd supports clusters of 3, 5, or 7 members for fault tolerance. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. Apache 2.0 licensed.