Your GitHub pull requests get reviewed by your own models, with inline findings posted in the merge flow: Codra is a self-hosted AI code review engine, so your diffs and credentials stay inside infrastructure you control. Connect a repository once and every opened, synchronized, ready-for-review, or reopened pull request starts a queued review job automatically; mention Codra for an on-demand pass instead. Findings arrive as inline GitHub comments, a summary review, and check run updates that fit the normal merge flow. Each repository gets its own review personality: triggers, skipped paths, draft handling, mention reviews, labels, custom rules, and a review budget are all tuned from the dashboard, and model routing follows global defaults, per-repository chains, fallbacks, and size-based overrides for larger pull requests. Providers from OpenAI, OpenRouter, Anthropic, Google, NVIDIA, and Cloudflare are managed behind GitHub OAuth sign-in. The operator dashboard tracks totals for reviews, input and output tokens, and comments posted, then drills into job history, webhook deliveries, failed queue runs, and model usage, so a stalled review is debuggable instead of invisible. Deploy the whole loop (GitHub App, web server, queue, database, and dashboard) on Cloudflare Workers, on Node.js 24, or with Docker on your own server. 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.

Real-time voice agents get a native Rust runtime in Flowcat: it carries a phone or WebRTC call through a composable media pipeline on your own infrastructure, so latency-sensitive audio never queues behind a managed platform. Audio flows from transport into voice activity detection and turn-taking, then through speech-to-text, an LLM, and text-to-speech, or a single speech-to-speech model, and back out to the caller, with in-process SIP and RTP for telephony. Pair it with local Whisper transcription, Kokoro, Piper, or XTTS for speech, and Ollama for reasoning, and a call's audio and transcript never leave your network; cloud providers plug in by name when you prefer them. Describe one agent in a YAML or JSON node-and-edge graph and the single flowcat-server binary serves it over HTTP, opening a browser page with microphone and live transcript, or bridging a Plivo number into the same agent. The runtime stays flat as call volume grows: benchmarked against a multiprocess Python baseline, one process holds p99 round-trip overhead at or below 0.61 milliseconds from 10 to 2,000 concurrent calls, at roughly 19.6 KB of RAM per idle session. Deploy it inside your VPC or an air-gapped environment where call recordings are sensitive. 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.

SealSkin is a self-hosted browser isolation platform that runs desktop applications in isolated containers on your server and streams them to any browser or phone, so links, files and downloads open remotely and nothing from the web ever executes on the device in front of you. Right-click a link, send a download, or drop a file from your local browser through the extension, and it opens in a fresh container on the server; close it and nothing is left behind, or keep a persistent home directory for work you return to. The catalog holds more than a hundred applications: browsers, office suites, IDEs, media editors, emulators and 3D tools from LinuxServer.io images, plus any Selkies-compatible image you add, with NVIDIA and DRI3 GPU acceleration for demanding workloads. Files stay on the server behind a file manager with chunked uploads, intercepted downloads, and password-protected public share links. Launch any app into a collaboration room with chat, voice and video, gamepad slots, and hand-over of mouse and keyboard control. Administrators brand the web app, manage users, groups, application keys, VPN profiles, and audit logs, and can route a session out through a WireGuard or OpenVPN exit at launch. Servers cluster with shared users and applications and no master node, and one manifest runs the whole thing on Kubernetes. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MPL-2.0 licensed.

E2A is an open-source email API that lets any application send transactional email and gives AI agents real two-way inboxes of their own, so an assistant can hold a conversation over email without borrowing a person's mailbox. Register an agent and it gets a working address on a shared domain in seconds, or verify your own domain for branded addresses, then connect through the hosted MCP server, a coding-agent plugin with browser OAuth, signed webhooks, a WebSocket stream, REST polling, or the TypeScript and Python SDKs. A local agent behind a firewall needs no public URL at all: it listens over WebSocket and replies in-thread with conversation identifiers keeping multi-turn exchanges coherent across humans and other agents. Every inbound message arrives with structured SPF, DKIM, and DMARC evidence, with the DMARC alignment verdict included, so your agent can judge how far to trust a sender instead of trusting the From header. Flip on the human-in-the-loop gate per agent and outbound messages pause for a person to approve, edit, or reject from the dashboard, the SDK, MCP tools, or magic links in their own inbox, with expiry rules and recipient allowlists narrowing which sends get held. Inbound content screening flags prompt-injection payloads before the agent ever sees them. Support bots triage intake, scheduling agents coordinate meetings, and commerce agents answer order questions from one persistent address. 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.

LynxDB is a single-binary log analytics database that ingests logs once into indexed columnar storage, so you can query the same events repeatedly without re-reading raw files or standing up a JVM stack. Pipe kubectl output or local files straight into the LynxFlow query engine for ad hoc investigation, or run the persistent server and let shippers deliver logs through Elasticsearch bulk, OpenTelemetry OTLP, Splunk HEC, syslog, and raw HTTP ingest paths. LynxFlow pipelines parse JSON, regex, key-value, CSV, and other formats at read time, then extend fields, filter with typed predicates, aggregate with stats, join CTEs, and rank results, all in one explicit grammar with time-series helpers for buckets, percentiles, gap filling, and trend lines. Full-text search runs on an FST dictionary, roaring bitmap postings, and bloom filters that skip cold segments, and materialized views keep partial aggregates ready so dashboards rewrite onto rollups automatically. Convert Sigma detections into LynxFlow queries for security hunting, tail live results as new logs arrive, and tier older data out to S3. Individual analysts use it to dissect local application logs, and platform teams run it as a cluster for shared observability with the Web UI and REST API open to everyone. 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.

OpenTickly is a self-hosted Toggl alternative that keeps your timers, projects, clients, and reports on infrastructure you control, so tracking time does not mean renting access to your own work history. Start the timer from the web app or an installable mobile PWA that works offline, then review the week in a color-coded calendar, switch to list or timesheet views, or open reports to see where hours actually went. Projects, clients, tags, billable rates, invoices, goals, and favorites organize entries the way agencies and freelancers already work, with team activity, top projects, and time-by-project panels on the admin overview. Because the public API mirrors Toggl Track v9 and Reports v3, existing scripts, integrations, and reporting jobs carry over, and coworkers can move their Toggl data in or back out without losing entries. The generous API also suits AI workflows: pair it with toggl-cli and an agent can track, organize, and review your time for you, then combine it with the Lyubishchev method to spot time waste across learning, social, rest, and work categories. Approvals, an audit log, integrations, and instance administration round out team operations, and seat limits, per-user fees, and usage caps simply do not exist here. 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.

The family calendar nobody opens helps nobody, so DinkyDash puts it on a screen the household walks past. Paste in iCal links from Google, iCloud and Outlook calendars and the board merges them into one agenda in time order, big enough to read from across the kitchen, with tomorrow tucked underneath when the day leaves room. A personal calendar can share only the events marked for a partner and keep work and private appointments to itself. Chores answer the daily argument on their own: whose turn it is moves on with the date, so nobody ticks anything off. Birthdays, holidays and special dates count down in the corner, and once a day the board gets a fresh headline written about your actual day, with a useful note parked under the agenda. Light and dark themes come from the settings page, and that same settings page on your phone edits the configuration file directly, comments preserved. A preview page shows the board at Pi, TV and tablet sizes side by side before you mount anything. There is deliberately no login: it assumes your home network, rewards you with a board you never sign into, and asks to stay behind a VPN or an authenticated proxy if you view it away from home. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.

xEdit is a free, open-source Markdown editor that keeps your notes as plain text, on your device or on your own server, and hands AI clients a proper door instead of a scrape. Write in live preview, where headings, code and images render in place and only the line under your cursor shows its source, or flip to a split view when you want both sides visible. Slash commands insert headings, callouts, tables, to-dos, code blocks and media, and wiki links tie notes together by title, with renames updating every link that points at the note. Any checkbox in any note with a date rolls up into a Today view, and nested folders, global search and a recoverable trash keep a growing library tidy. Choose where notes live: browser storage, a local folder of Markdown files, or cloud sync with an offline mirror on every device. Cloud notes gain version history with line-by-line diffs and one-click rollback, share links that collect comments on selected passages, and exports to Markdown, HTML, PDF, PNG and Word. Pick one of thirteen layout themes and copy straight into WeChat or Zhihu with styles inlined and math intact, then connect Claude or Cursor through the built-in MCP server to search, organize and edit the whole library. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.

Garcon gives developers one browser workspace for running several coding agents at once, steering their work live, and turning reviewed changes into commits. Tile up to four resizable windows, move chats between them, and place live sessions on durable canvases that show delegated children, handoffs and parentage. Turn that activity into a tag-driven delivery board, or create tickets first and open chats from them with the project, priority, assignee and discussion already attached. Queue instructions, reorder them, answer agent questions, approve tool calls, pause work, interrupt a run or stop it without losing the transcript. Exchange labelled messages between chats, delegate bounded tasks to child agents, and bring final results back into the parent conversation. Browse and edit files, open terminals, inspect reasoning and tool calls, then stage individual lines, hunks, files or folders before committing and pushing. Search transcripts, save filters, pin or archive chats, export Markdown or XML, schedule prompts for later, and review token use and estimated API cost by day, agent and model. Connect trusted remote executors when work belongs on another machine, and open the same workspace from a phone when an agent needs a decision away from your desk. 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.

Your app needs avatars, attachments and user photos, and imgpush is the tiny service that takes them off your hands. Send one upload request with a file or a URL and you get back a filename; ask for that file with a width and height in the address and you receive exactly that crop, converted to the format you configured. Originals live on local disk or in any S3-compatible bucket (AWS S3, Cloudflare R2, MinIO or Backblaze B2), with resized variants cached locally for fast serving. Abuse controls come built in: per-IP upload quotas per minute, hour and day, an allowed-origins whitelist, and an optional nudity filter that can also sample video frames before an upload is accepted. An API key can be required for uploads, deletes remove the original together with every cached size, and a liveness endpoint keeps Docker and Kubernetes health checks honest. One Docker container on port 5000, a page of environment variables, and your application never writes image code again. 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 sales team's conversations scatter across a dozen apps, and HiveMTK pulls them into one private workspace where AI agents do the daily chasing. Douyin, Xiaohongshu, TikTok, WeChat, Telegram, WhatsApp, email and SMS all land in a unified inbox, and every customer merges into a single profile no matter where they first appeared. A ReAct agent perceives, plans, calls tools and reflects instead of following a fixed script, composing its 45 tools across outreach, customers, knowledge and business tasks to handle situations nobody scripted. A proactive partner agent works the other direction, mining group leads and starting conversations, and a visual canvas turns the whole playbook into marketing flows anyone on the team can edit. The knowledge base answers from your own documents with hybrid vector and keyword retrieval, and because inference runs locally, chats, embeddings and customer data never leave your network. Dashboards track leads, conversions and channel mix, and a GEO module works on getting your brand cited by AI search. It is a full acquisition to repurchase funnel for teams of five to fifty who refuse SaaS lock-in. 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.

Lawyers work in piles of text, and AI WorkDeck gives them the IDE that programmers take for granted: every matter keeps its files, drafts, evidence and AI help in one window. The AI edits the real document, not a chat transcript, writing its changes straight into the file as tracked changes that you accept or reject one by one. Every conclusion carries a citation, and clicking it lands on the supporting passage in the source document, with changed or missing anchors flagged for review. Drafts live under git-backed history with three-way merges and paragraph-level provenance, so two people can revise the same pleading and still see exactly who wrote what. Out in the field, photos and recordings taken on a phone land back in the matter on your computer, filed by date and ready as evidence. A built-in marketplace adds one-click Skills, plugins and MCP servers, and a Word add-in meets colleagues who never leave their usual editor. Run it as a desktop app for one lawyer or as a server for the whole firm, where accounts, sync and shared matters live on infrastructure you control. 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.

Point an AI pentesting crew at a web app you own and watch it hunt: BugTraceAI plans the checks, sends specialist agents after each weakness class, and brings back evidence you can inspect. A scan moves through six visible phases, from crawling and endpoint discovery through strategy, exploitation and validation, and ends in a findings table where every result carries a severity, a CVSS score and a confirmed status. The web dashboard follows the hunt live as a graph of specialists branching from a coordinating agent, and its API discovery workspace brute-forces routes and loads the interesting ones straight into a scan. Analysts get their own toolbox of more than twenty utilities for poking HTTP traffic by hand, an AI repeater for replaying requests, and ModelLab for comparing models on cost and results. Findings download as a ZIP with the full report, and a terminal workspace with pipeline, findings, agents, timeline and logs views suits keyboard-driven work. Bring your own model through OpenRouter, Anthropic, Z.ai or a local endpoint, and aim the whole suite at the bundled BugStore practice shop when you want to rehearse safely. 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.

One dashboard and one API address for every local model you run, across llama.cpp, vLLM and MLX: that is llamactl. Create an instance in the web UI, pick a backend, and the server assigns ports and API keys, starts the model, and shows its health next to every other instance you run. Models download straight from HuggingFace inside the dashboard, in GGUF for llama.cpp or Safetensors for vLLM, so stocking a new model never means juggling command lines and file paths. Your applications keep talking to familiar endpoints, because llamactl exposes OpenAI chat completions and Anthropic messages routes and sends each request to the model named in the call. A llama.cpp router instance can serve several models from one process and load them on demand, and idle instances stop on a timer or make room through LRU eviction when the GPU gets crowded. Groups cap how many heavy models run at once, backends can live in Docker containers with GPU access, and instances on other machines appear in the same central dashboard with automatic routing. API keys minted in the UI guard both inference and management calls. 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 own video hosting site that behaves like the big platforms: upload a file, share a link, and anyone can press play. Subatic takes care of everything between those two moments. Drop in a raw video and the transcoding pipeline converts it into web-ready formats, stores the finished files in S3-compatible storage, and publishes a clean watch page with the title and an animated thumbnail that previews the content before a click. Playback streams directly from the storage bucket, so a Cloudflare R2 bucket with caching turns your little server into a video host whose delivery costs stay close to zero as views climb. A public homepage grid showcases your latest uploads with those moving thumbnails, and visitor counts can flow into Umami or Plausible if you want audience numbers without handing viewers to a third-party tracker. Creator accounts handle uploading, and one Docker Compose file brings up the whole stack: the app, the transcoder, MinIO for local object storage and PostgreSQL. A teacher can post course videos without platform rules, a band can share rehearsal footage, and a homelab can keep family clips on hardware in the closet. 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.

Always-on AI teammates that each get a real computer of their own, with a browser, files and a terminal, working from your hardware instead of a vendor cloud: that is OpenSuperBot. You hire a teammate from the sidebar with a name and a one-line job, and it works inside its own sealed computer with a browser, shell and persistent workspace. At a login wall the bot asks you to take over its live screen and sign in once inside its own browser profile, so passwords never pass through the gateway. Group threads let you ask the whole room at once: everyone answers in parallel, the chief of staff closes with a dispatch table, and @names pick which teammates respond. Nothing leaves the workspace without your say-so, because outward actions stop at a draft-and-hold approval chip. Standing approvals cover repeat sends to one destination for an hour or a day, and money and bookings always ask. Memory is a visible file you control: tell a bot a rule and it posts the exact change as a chip, and forget removes it the same way. Routines come from plain conversation, so a daily digest, a secret webhook trigger or a quiet status-page watch can be scheduled in chat. Each teammate runs a model you pick, with daily token budgets and live step cards for every action. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.

DopplerTask is the control layer where AI agents build your automations and you decide what goes live. You describe the outcome in plain language, and your agent assembles a task from a library of eighty-one ready-made steps spanning messaging, ticketing, sales, operations, documents, databases, queues, and AI models. The proposal lands in a visual editor that people and agents share, with new and changed steps marked and the old value of each field beside the new one. Nothing goes live until you approve it, and the previously approved version keeps running untouched in the meantime. Once a task is live it can start from a webhook, a schedule, a message queue, a manual click, or the agent itself, and every run is recorded with the output of each step, the person who started it, and the person who approved the version. Failed runs resume from the step that failed, with retries and continue-on-failure where you want them. Agents reference your credentials by name only; the values stay encrypted on your server, beyond the agent's reach, and separate tokens for people and agents plus single sign-on keep access tidy. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.

Describe the change in chat, and LoopForge opens a Docker sandbox, checks out your repositories, hands the work to specialist agent roles, and opens the pull request. That is the conversational path: a Main Bot coordinates planning, implementation, review, and testing roles, each with its own prompt, executor, model, rules, skills, and tools, and you can queue or steer messages mid-reply, then pause, resume, or take over a run at any point. When a way of working proves itself, you freeze it as an SOP: a visual canvas of agent steps, deterministic script steps, human gates, and sub-workflows that every requirement then follows in the same order with the same checks. Four built-in SOPs cover backend delivery, frontend delivery, product proposals, and bug fixes, and each one can be copied, edited, and versioned per project. A Needs review inbox collects every gate waiting on you, with per-node status, tool traces, logs, and artifact snapshots as the evidence behind each approval. Recurring AI tasks run on schedules with the same workspace and workflow access, and insights reports track engineering metrics and token costs per run and per conversation. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.