Nakama gives your whole team a shared home for AI agents, where each agent has its own identity, memory, tools, and permissions and meets people where they already talk. The dashboard feels like a familiar chat app: pick an agent from the rail, start a conversation, and watch it use tools, recall memory, and reply inside Telegram, WhatsApp, Discord, or Slack when the work lives there. Every agent is configured as a small personality you can read and edit, with prompt files for its soul, style, instructions, and memory, plus tabs for its channels, tools, MCP servers, and skills. One server hosts many organizations with invites and roles, and agents stay isolated from each other so the support agent never wanders into the finance agent's toolbox. The control center manages workspace settings, model providers, plugins, and integrations, and a connector layer opens the door to more than a thousand external apps. Daily encrypted backups and an in-app update button keep operations boring, and the entire platform idles in about two hundred megabytes inside a single container. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.

Aurora is an open-source AI SRE agent that investigates production incidents on its own, turning each incoming alert into a structured root cause analysis with evidence and a suggested fix. Alerts arrive from PagerDuty, Datadog, Grafana, New Relic, OpsGenie, incident.io, or CloudWatch, and every one triggers a background investigation without waiting for a human to pick it up. The agents choose from more than thirty tools, run cloud and cluster commands inside sandboxed pods, query logs, compare deployments, and search a knowledge base of past incidents for similar failures. An infrastructure dependency graph shows the blast radius across services and providers. When the root cause is found in code, Aurora can draft the remediation as a pull request instead of stopping at a diagnosis. Finished investigations become postmortems with a timeline, impact assessment, and remediation steps, exportable to Confluence, Notion, or SharePoint, and automated actions can open the fix, notify Slack, or update living documents as new evidence lands. Command policies and guardrails bound what the agents may run, and any major language model provider can power the reasoning, cloud-hosted or local. 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.

Plenty of sites you follow have no feed, a broken feed, or announcements buried in an inbox, and mkfd turns all three into clean RSS feeds you control. You point it at a webpage, a REST API, or an IMAP mailbox and describe what counts as a story: an item selector plus selectors for title, link, date, author, and enclosures. Drill chains follow links across several pages when the real content sits deeper than the listing page, and an advanced scraping mode with a real browser handles pages that need JavaScript or your own browser extensions. An integrated selector helper helps you pinpoint elements on the target page instead of guessing. Email feeds watch a folder continuously and publish new messages as feed items the moment they arrive, with mailbox passwords encrypted before they touch disk. Every feed gets a schedule, is rebuilt in its own worker, and is served from an endpoint your reader can subscribe to, and the whole configuration lives in a portable file you can version and reuse. Access to the builder is guarded by a passkey, and feed endpoints stay open for your reader. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.

Turn your spare servers into a Discord bot hosting business: Mechon gives your customers a clean panel while every bot runs locked inside its own sandbox. You define the plans you sell, pools of bots, memory, CPU, and disk, with per-customer overrides when a good client needs more. Customers create a bot from a discord.js, discord.py, or Bun template, paste a token, and deploy by uploading a folder, pushing from git, or shipping from CI. Every bot runs as its own Linux user in a locked-down container with a read-only filesystem and kernel-enforced limits on memory, CPU, processes, and disk, so one customer's runaway bot cannot read another's token or starve its neighbors. The live console streams logs as they happen, with memory, CPU, and disk usage per bot and capacity per server. Deploy history records every release and rolls back the bad ones. An API and signed webhooks connect the panel to Paymenter, WHMCS, or Stripe, and the operator keeps the servers, the customers, and all of the money. Server nodes run one agent next to Docker and dial out to the panel, so no inbound ports need opening on your machines. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.

AI Native Call Center is a self-hosted call center where a voice model answers the phone first and people step in only when the conversation actually needs one. Calls land on FreeSWITCH and connect straight to a speech-to-speech provider, and a flow engine steers each call through phases that carry instructions and the tools the model may use, with transitions firing on tool results. When a person is required, the call transfers into a real queue and the agent's browser screen has already popped with the caller's details and the bot-phase transcript ready to read. Agents handle presence, call control, and callbacks from the web interface, and a companion browser extension holds their SIP registration and audio. Supervisors watch a live wallboard with the queue view and the agent roster. Persona, rules, and voice are published and versioned together, and flows are plain documents you can author and validate before loading. Every conversation, bot-answered, handed over, and finished by an agent, lands in a single call record holding one transcript and one recording stored on a filesystem or an S3-compatible bucket. The interface and the bot work in English and Chinese, with provider choices for different regions including an option to run your own realtime gateway. 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.

When a client needs a two-gigabyte video and your options are a consumer transfer site with a weekly cap or an FTP account older than your interns, FileSharing gives you a private WeTransfer of your own. You group files into a bundle, drag them in from your filesystem, and keep adding more days later if the project grows. Every bundle produces two links: a preview link where recipients see the file list first, and a direct link that downloads everything at once, both guarded by an authorization code. Bundle settings cover a title, a description, an expiration date, a maximum download count, and an optional password that also encrypts the ZIP archive recipients can pull down. Upload access stays yours to control through an IP allowlist, login and password accounts, or a mix of both. Optional duplicate prevention skips files you already sent, a download rate limiter keeps a single recipient from saturating your bandwidth, and a scheduled background task physically deletes expired bundles so old transfers do not pile up. There is no database to run or back up because bundle metadata lives in JSON files on disk, and the interface speaks English, French, German, and Korean. 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.

Cotal puts your AI agents in one shared room: Claude Code, Codex, and OpenCode discover each other, message channels, hand off work, and ask you to approve the result. Post to a channel and every subscriber hears it; message one peer and the durable inbox holds it until they are free; address a role like reviewer and one available reviewer claims the job. Presence shows who is actually there, and attention modes decide which channel traffic may interrupt an agent mid-task. A manager starts and stops agent processes in tmux or cmux, here or on remote hosts you authorize, without sitting between their messages. For work that waits and branches, Cotal Lang programs launch agents, pause at human checkpoints with deadlines, and resume from their recorded steps after an interruption. Services join the same space by publishing command contracts that authorized agents can discover and invoke; the bundled Linear adapter shows the pattern. Supervise from a terminal console with the live roster and activity feed, or from the read-only browser dashboard and its graph of agents and channels. Spaces stay separate, every message carries a checked identity, and per-agent grants decide who may delegate, spawn teammates, or call a service. 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.

Open Walnut replaces your scattered terminals and notes with one self-hosted command center where a CEO agent and a lead for each project run your Claude Code sessions like a small company. You talk to the CEO; each project lead plans the work, spawns worker sessions, checks their output, and only brings you the questions that actually need a person. Every session sits in its own column with live output, permission controls, terminals, and files, and sessions survive browser disconnects, dropped SSH connections, and server restarts. Codex, Gemini, OpenCode, and Goose sessions join the same columns, on your laptop or on remote dev boxes reached over SSH. Finished work comes back as a per-session diff: comment on individual lines and send the review straight back into the session that wrote it. Around the agents sit your tasks with subtasks, dependencies, and due dates, Markdown notes with wiki links and backlinks, a calendar, and long-term memory that a background job keeps consolidated. Search covers notes, tasks, sessions, and memory together, mixing keyword matching with local embeddings, so nothing you told it last month gets lost. Routines run one-time, interval, and cron jobs; plugins add whole apps, tools, and skills; and paired phones reach the same console through device tokens. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.

Hexis gives every AI agent in your company one governed brain: skills, tool manuals, and knowledge stored as files in a git repository you own. Engineers and non-technical teammates browse the same workspace to load plugins, read company context, and propose improvements without touching git themselves. A proposed change shows its exact diff to the owners of the files it touches, and it ships only when they approve, on the branch rules you set. Agents propose too: one that hits a broken skill mid-task can suggest the fix as a change request and wait for a person to decide. Claude Code, Cursor, Cline, Codex, and ChatGPT connect over MCP and discover skills by search at call time, so the context window stays lean. Each agent acts with its user's own permissions, resolved per file, and the credentials behind tools stay in an encrypted vault the agent never holds. Roles and folder-level grants decide who sees what, and your team signs in behind your own SSO. Every save is a commit, so history, blame, and revert come free, and the knowledge base can start from the bundled template in minutes. 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.

Your folders are the library with MediaServer, a self-hosted player that scans the directories you already keep and serves their video and audio through a tidy web interface. Browsing feels like a personal YouTube: folders open as shelves of content, and the custom player keeps the video going as you explore what to watch next. Keyboard shortcuts drive playback, speed controls and player statistics sit a click away, and an ambient background glows behind the video. Every view is remembered: full watch history with timestamps, re-watch counts, per-user totals, and playback heatmaps replace the single watched tick other servers stop at. Metadata belongs to you: edit titles, thumbnails, tags, seasons, and release dates for videos, folders, and albums, and the edits survive files being moved or renamed. Music gets its own treatment with embedded cover art, a lyrics viewer and editor fed by LrcLib, and auto-scrolling lines that follow the song. Subtitles extract themselves from embedded and external files, folders and single videos share as readable links, and each share renders a rich preview card for Discord or chat. Behind the scenes, a dashboard manages libraries, scanning jobs, users, and server performance. 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.

Apple Music, NetEase Cloud, and QQ Music stop being separate apps with MHub: one self-hosted player searches all three and lines their songs up in a single queue. Search any platform, import its playlists, and MHub matches tracks across services, then keeps imported lists following the originals with a daily sync. The queue takes appends, play-next picks, drag reordering, shuffle, and loop, and an AutoPlay radio built from your recent songs keeps playing when it runs dry. ALAC tracks decode losslessly right in the browser with sample-accurate scheduling, and everything else plays natively, with quality chosen at play time. Lyrics land character by character over a background tinted by the cover art, and they convert between Simplified and Traditional Chinese on request. Lock the phone or switch apps and playback rolls on with system media controls, the next track preloaded for a seamless handoff. Songs and lyrics cache on the device for offline listening, and the player installs as a PWA that behaves like a native app. Add a passkey account and your library syncs live between devices, so the playlist you built at your desk is waiting on your phone with your place kept. One Docker container runs it, bound to localhost by default behind your reverse proxy. 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.

Everyone on your Plex server gets their own private recommendation rows with Shortlist, built from what each person actually watched and visible to nobody else. For each viewer it reads their Plex watch history, picks unwatched titles from your library, and explains every pick with the reason it earned its place. Rows land on Plex Home labelled, hidden from every other account by share filters, and only then promoted, so a row is never visible to the wrong person, not even briefly. Six row kinds, from Picked for You to seasonal picks and personal requests, start from eleven templates instead of a blank form. A seasonal row follows the calendar through Halloween, Christmas, and twenty more ready-made occasions, then hides itself between seasons. Rebuilds run nightly, weekly, or monthly on your cadence, and the dashboard shows what was delivered against what people started and actually finished, per user and per row. Missing a strong pick, Shortlist asks Radarr, Sonarr, or Overseerr for it: the boldest few auto-send nightly, the rest wait for one click in the Requests inbox. No AI key is required; add one and Describe a row turns slow-burn heist films into a themed row filled nightly for each person. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.

The audit starts with one URL: Open SEO Crawler spiders the entire site from your own server and groups every technical issue by severity. It crawls concurrently with per-host politeness and backoff retries, and presets adapt automatically to WordPress, Shopify, Webflow, and seven other detected CMS platforms. Every page gets checked for missing or duplicate titles, meta descriptions, H1s, and canonicals, plus thin content, redirect chains, broken links, hreflang gaps, and images without alt text. Afterwards, sample pages are fetched again as Googlebot, Bingbot, Applebot, GPTBot, and PerplexityBot, and any page a bot cannot read in full gets flagged. The Summary groups issues into errors, warnings, and info with affected page counts, and the sidebar opens bulk reports for titles, metas, canonicals, duplicates, response codes, and deep pages. Each report exports to an XLSX workbook, and the Crawl Budget page turns parameter traps into ready-to-paste robots.txt rules that stop phantom URLs from eating your budget. One click packs the crawl into a Markdown brief for any AI assistant, and the bundled MCP server lets Claude or Cursor start crawls and read issues directly. Saved crawls reopen any time, dark mode is built in, and a daily auto-update keeps the tool current with rollback if a pull ever breaks it. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.

Your couch becomes the control room with RelayTV, a self-hosted media hub that runs on the Linux box plugged into your TV. Connect Plex, Jellyfin, or Emby once, then search films, open a series, pick an episode, resume where you stopped, or drop the pick into the shared queue. Paste a web video link to play it on the big screen, or queue an entire YouTube playlist of up to a thousand videos in order without pausing tonight's movie. The TV idles on a dashboard with a QR code, and anyone in the room scans it to join the same remote, browse, and add picks to the queue. Add M3U playlists to build a My Channels lineup, pin favorites to the top, rewind inside the buffered window when you miss a line, and jump straight back to live. Playback, queue advancement, history, and overlays all run on the box itself, so the TV keeps playing after your phone leaves the page. Every screen appears in Home Assistant as a media player, and the local HTTP API takes play, queue, overlay, and snapshot calls from scripts and dashboards. A second RelayTV in the bedroom can receive the active queue from the living room, and the optional API token locks control down on shared networks. 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.

Myriad gathers your games, music, anime, and social posts from eleven platforms into one self-hosted homepage that has a resident character of its own. Connect GitHub, Steam, Bilibili, NetEase Cloud Music, and seven more platforms, and every account lands in one database that the report cards and library read. Report cards, a music player, weather, friend links, and recent activity fill a homepage you arrange on a grid or a free sticker canvas. The library shelves the collections as games, anime, films, books, and music, and clicking a report card plays a three-act portrait: your persona there, your taste, and your highlights. Her layered 2.5D portrait blinks, mouths each reply, and shifts expression in the browser, and a long press opens her panel anywhere on the site. In Chat she stays in character; in Help she plans, asks you to confirm, then acts inside the permissions you granted. The studio turns notes into a static Hugo blog, and the built-in reader follows your RSS, Notion, and RSSHub subscriptions. Tapps run sandboxed as pages or homepage widgets, and the Playground previews one from a single typed sentence before you install or export it. Follow other sites over ActivityPub, pair QQ, Telegram, or Discord bots that answer mentions in group chats, and keep registration and module visibility under your own switches. 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.

Every animal gets a proper health record with EinVault, a self-hosted tracker for companion care covering vet history, daily routines, and the helpers involved. Each profile holds species, breed, vet details, emergency contacts, and a photo, and the activity choices adapt to the animal, from dog walks to litter box visits. A daily journal records mood, photos, and videos, and tagged markdown notes keep feeding instructions and quirks pinned where caretakers find them. Health tracking covers vet visits, vaccinations, procedures, and weight history; medications carry doses and end dates linked to scheduling reminders. A one-tap Given log stamps each dose and shows the next due time, and incidents follow an injury or symptom to resolution with severity, updates, and owner alerts. For time away, a printable caretaker sheet puts schedules, contacts, and selected notes on one page for a sitter who needs no account. Recurring reminders, a calendar feed, full-text search with companion and type filters, and a token API for logging from smart buttons or scripts round out daily care. One Docker container runs it with admin, member, and caretaker roles, optional single sign-on and two-factor authentication, and six interface languages. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.

Draw a pipeline, push code, and watch it build, deploy, and roll back on its own: Pipewright is a self-hosted CI/CD platform delivered as one static binary. Pipelines are drawn on a two-level canvas of stages and jobs, then run with matrix builds, sidecar databases for tests, manual approval gates, and typed parameters validated at trigger time. The pipeline can also live in the repository as a per-branch .pipewright.yml file, and triggers cover webhooks, schedules, branch to environment mapping, and chained pipelines. Builds run isolated inside containers with cached dependencies and a content-addressed artifact store, and JUnit and Cobertura reports feed quality gates that stop bad code before deployment. Deployment is agentless over SSH: health checks gate a zero-downtime cutover, failures roll back on their own, and environments track their active version with one-click rollback and promotion toward production. Deployed services receive HTTPS domains through a managed Caddy proxy with automatic certificates, path routing, and load balancing, and each pull request gets a throwaway preview address with its own certificate. The operations side watches host CPU, memory, and disk, manages containers and networks across servers, and opens terminals in the browser. Notifications route by event to Slack, email, WeCom, DingTalk, Lark, or a webhook, approvals arrive as signed links, and the four DORA metrics are computed out of the box. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.

Denpa is a home TV recording server that turns a Linux machine and a TV tuner into a private recorder for Japanese terrestrial and satellite broadcasts, with reservations and browser playback. There are no configuration files: plug in a supported tuner and card reader, press scan, and the tuner type, channels, and program guide set themselves up. Reserve from the guide, or save a search as a rule and matching broadcasts record automatically; a repeated episode is ignored after its first airing is captured. Recordings encode to AV1 or H.264 MKV files (H.265 when a GPU is present) with captions exactly as broadcast, and detected commercials become chapters that playback skips entirely. Playback resumes where you stopped, runs at up to double speed, and clips save on the server with captions intact; programs also download to devices for offline viewing. Live viewing runs about a second behind the broadcast: pausing starts chase playback, alternate audio stays selectable, and BML data broadcasting works in the browser. The companion denpa-tv app reaches Android TV and Fire TV through QR pairing, and an API exposes channels, recordings, and audio-only feeds to Home Assistant and scripts. EPGStation and EDCB users can import rules, reservations, and recordings, and the interface stays closed until you open it to trusted networks or protect it with OIDC login. 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.