Hivemill
When you hand Hivemill a goal, AI coding agents build it, separate agents review and test it, and a judge decides the merge, all on a Linux server you control. You describe the outcome to a boss agent, a director staffs a lead, and the lead splits it into tasks, each with its own git worktree and branch. A coder opens a pull request and stops, because it cannot review, test, or merge its own work. The reviewer reads the diff, QA runs real tests against your acceptance criteria, and by default the judge merges only when both verdicts pass; anything else returns to the coder with findings. Pull requests, CI checks, and merges run through the GitHub CLI. Claude Code is the default runtime, a Cursor Agent org can run beside it, and hooks enforce each role's prompt and permissions. A spawn gate queues new agents until memory, load, and your Claude usage windows allow, so one modest machine stays responsive. Scheduled jobs mail the director recurring chores with a token cap per run, and a daily retro agent studies rejected work. Typical first jobs are bounded changes such as a filtered CSV export, a mobile navigation fix, or a duplicate-order check before import. A read-only web dashboard tracks goals, tasks, live agents, throughput, quality, and estimated cost per task. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.
Clawboo
Describe the outcome once, in plain language, and a small crew of AI agents divides the work between them and reports back in the same chat. You type the request in the team chat; Boo Zero, the team lead, splits it into jobs on a shared board and hands each card to the teammate best placed to do it. Teammates run on Claude Code, Codex, Hermes, OpenClaw, or Clawboo's built-in runtime, and one team can mix them freely, so the researcher, the writer, and the coder can each work on their strongest tool. You follow every card live on the board and in the chat, watch the team graph to see who is doing what, and add a reviewer agent when a second pair of eyes matters. Finished work returns to Boo Zero, which pulls the pieces together and answers you in the chat. Ready-made teams from the marketplace cover marketing squads, dev teams, and research crews, and you can bring an API key from Anthropic, OpenAI, OpenRouter and more, a free local model through Ollama, or your ChatGPT subscription through the Codex CLI. Spend is tracked per team with budget warnings before costs surprise you. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.
MATE
Change an agent's prompt, model, or audience from a dashboard and watch it go live without a single redeploy. MATE opens in a Work Room where any root agent can be picked from a card grid and chatted with directly. The Studio holds the builder: a drag-and-drop canvas draws parent and child agents, attaches tools and MCP servers, and keeps every configuration versioned with one-click rollback. Trace Replay plays a past conversation back across that canvas so each hand-off is visible. A template library imports ready-made agent teams, from research pipelines to support desks. The Control Room covers governance: role-based access per agent, dollar costs per agent and project, budgets that warn or block, guardrails for personal data and prompt injection, and alerts to Slack, Discord, email, or a webhook. Eval suites score replies by exact match, similarity, or an LLM judge, and a Playground compares models or prompts side by side on score, speed, and cost. Agents can be embedded as a chat widget, called through an OpenAI-compatible endpoint, or exposed over MCP and A2A. Agencies running client agents and platform teams juggling dozens of them can manage the lot from one dashboard. 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.
Trinity
Trinity runs your AI agents the way a company runs employees, with departments, roles, reporting lines, and receipts for everything they do. Each agent holds a seat on record tied to an objective and a declared metric, so whether it is working is a number with a date on it rather than an impression. Agents work on three runtimes, Claude Code, OpenAI Codex, and Gemini CLI, chosen per chat without changing the agent's state, history, or permissions. A schedule fires, the agent works inside its own container, and anything irreversible waits in a human approver's Inbox as a card naming the agent, the skill, and the input until someone approves, discusses, or rejects it. Every agent is a git repository you own: it commits its own work, its tool calls land in an append-only audit log, and every run appears in an execution ledger with its trigger, cost, and outcome, searchable across the fleet. The same agents answer in the browser Workspace, in Slack, Telegram, and WhatsApp, over a public chat link, through MCP, and from the CLI. Guardrails bind agents by what they hold rather than what they are told, with hooks they cannot switch off, cost limits, loop budgets, and a kill switch for outbound calls. 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.
Optio
One feed holds every agent job at once: pull requests in progress, nightly briefings, incident investigations, and the terminal you left open on a laptop. Optio gives Claude Code, Codex, Copilot, Gemini, Cursor, OpenCode, Pi, or OpenClaw a shared workspace, so the work stops scattering across tabs. A short form answers five questions, when, where, who, what, and then, and reads the result back as a plain sentence. Triggers include schedules, webhooks, and events from GitHub, GitLab, Slack, Linear, Jira, PagerDuty, Sentry, Alertmanager, and Datadog. Repository tasks run in git worktrees, open pull requests, follow CI and review, and resume with feedback. Sessions open as conversations with a terminal beside them, and persistent agents keep their own inbox and instructions. A Needs You queue and provider usage limits sit on the overview page, and native iOS and Android apps put the queue, widgets, and Live Activities on a phone. Connections store encrypted credentials and MCP tools per repo and runtime. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.
CompozyOS
CompozyOS keeps your Claude Code, Codex, and Cursor agent sessions alive after the terminal closes and sorts them on one board by who owns the next step. Sessions land in Needs you, Working, Waiting, In review, or Done. Parallel agents each get their own worktree. The branch, the pull request, and its checks sit on the session. Fork a session, rewind it to a checkpoint, steer it mid-turn, or hand it to another agent CLI. Approvals hold a tool call until a person allows it. Always-on agents check in on a schedule, and a quiet check costs no model call. They watch sources such as new GitHub issues and propose actions. Approve each proposal, or let the agent act alone up to a daily limit. Loops pair a goal with a verifier and an iteration cap. Every Loop ends in one of seven named outcomes. Schedules, runtime triggers, and signed webhooks start agents on their own, and every run is recorded. Fifteen built-in integrations cover ten agent CLIs and five providers, on your own keys. A Marketplace adds extensions, plugins, and MCP servers in one step. The daemon starts reachable only from the machine itself, and remote access stays off until you turn it on. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.
pi-dispatch
pi-dispatch turns one AI coding agent into a swarm that works in the background and reports to a single web page. Jobs start from a nightly schedule, a labeled issue, or a comment, then wait in one shared queue until a machine has room. Each job runs in a sealed box that is thrown away afterwards, reaches only the internet addresses you allow, and stops at the dollar limits you set per day, week, and month. When an agent hits a decision, it parks a question with options and a recommendation, and no model spend accrues until you answer. The overview shows what needs you, what is queued, what is running, and what today has cost. A Zero touch page tracks how often runs needed a person, per project, so you can see how far to step back. On GitHub, GitLab, Forgejo, and Azure DevOps the work ends as a proposed change that a human merges. Fleet, Spend, Triggers, and Models pages cover several machines, budgets per project, and fallback models. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.
LiteLLM Agent Control Plane
Agents scattered across vendor consoles become one tidy workspace in LiteLLM Agent Control Plane, which puts Claude Managed Agents, Cursor, OpenCode, OpenClaw, Deep Agents, and Hermes behind a single login, one API, and one dashboard. Describe an agent in plain language or start from a template for a deep researcher, an inbox triager, or a security reviewer, then tick the tools, skills, and MCP integrations it may use and pick the runtime that executes it. Chat happens inline with streamed output, sessions persist across runs, and a memory layer carries context from one session to the next. Routines launch agents on a cron schedule, and each scheduled run lands in the history as its own session. An MCP proxy injects upstream credentials at the wire, so agents never hold real secrets, and provider keys sit encrypted in a vault. Colleagues can reach the same agents from Slack, Microsoft Teams, Google Chat, or a webhook, and developers can drive everything through a REST API with a live try-it reference. Administrators issue per-user virtual keys with budgets, model limits, and expiry dates, and an inbox gathers anything that needs a human decision. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.
Grasp
One sentence in, one reviewed pull request out: Grasp sends coding agents through a visual workflow in isolated sandboxes and leaves every approval to people. Describe a task on the home page and run it as a workflow or a Solo session. A clarify agent first asks one question at a time as clickable option cards with a recommended answer, then writes a requirement spec and plan and builds a page mock you can annotate. Workflows are drawn on a visual canvas with six node types, and failure and rollback edges send work back to an earlier step. Eight agent CLIs, including Claude Code, Codex, Cursor, and OpenCode, can run as agent nodes in a Docker, Kubernetes, or E2B sandbox shared by the nodes of a run. Runs pause at human gates in a shared inbox, and a temporary one-time link lets a reviewer without an account approve or reject. Solo sessions give a single agent an IDE, terminal, and browser, and a noVNC panel lets you watch its Chromium and take over with a click. Specs, plans, test results, and HTML previews are kept as versioned artifacts. A token page breaks cost down by workflow, source, and model. Agents push branches and open pull requests or merge requests through gh, glab, or SSH credentials stored per repository. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.
Hezo
A whole company of AI agents, from a CEO to engineers, designers, and analysts, works inside Hezo under budgets, approvals, and sandboxes you control. Starting a project means picking a ready-made marketplace team (App Team, Social Media Marketing, or Investment Portfolio), using a template, or asking the CEO to assemble one, and the org chart can be reshaped while work continues. Agents pull tasks from a board with nested sub-tasks, wake on heartbeats, pause at approval gates, and push verified commits to your GitHub repositories. Each agent can run on its own model, whether Claude, GPT, Gemini, Grok, Kimi, DeepSeek, GLM, OpenRouter, Ollama, or LM Studio, using API keys or an Anthropic or OpenAI subscription signed in from the UI. Token budgets pause runs at their limits, cost per run stays visible, and a container-hours ledger enforces a monthly allowance. Every project works in its own container, and stored secrets reach agents only as placeholders that an egress proxy swaps for real values on hosts you allowed. A Coach agent turns lessons from finished tasks into durable rules, while documents, skills, and full-text search hold shared knowledge. Message the CEO from Telegram, Slack, or Discord, or drive Hezo from any MCP client through its built-in MCP server. The web app speaks 12 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-or-later licensed.
Archestra
Archestra delivers the enterprise AI infrastructure layer that organizations need when managing multiple LLM providers, MCP servers, and AI agents across teams becomes unmanageable. The LLM gateway routes requests across Anthropic, OpenAI, Azure, Bedrock, and DeepSeek with virtual API keys, per-team cost limits, and dynamic model routing — giving every developer one token for Claude Code, Cursor, or Codex while finance tracks spend per department. The MCP gateway authenticates tool calls with OAuth 2.1 and On-Behalf-Of tokens so each tool executes as the calling user, not a shared service account, eliminating credential sprawl. The private MCP registry lets teams publish custom tool servers with approval flows promoting servers from dev through staging to production, each environment maintaining its own credentials and network egress policies. The Kubernetes operator manages MCP server lifecycle — deploying containers, scaling, health-checking, and routing gateway traffic to local servers automatically. The agent runtime supports scheduled triggers, email and webhook invocations, sub-agent delegation, reusable skills, and sandboxed code execution with a K8s-native filesystem. Deterministic guardrails including Dual-LLM verification and Lethal Trifecta protections prevent dangerous tool calls before execution. Built-in OpenTelemetry traces and Prometheus metrics provide full observability without additional tooling. Docker deployment exposes the Admin UI on port 3000 and API on port 9000 with a single command. 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.
Edict
Edict is a multi-agent orchestrator that runs every AI request through a court-style chain: a crown prince agent triages it, a planning department drafts the work, a review department can veto the plan, and specialist ministries execute what survives. Casual chat gets a direct reply, while real commands become kanban tickets whose titles are scrubbed of file paths and metadata. A vetoed plan returns for rework; an approved one is dispatched in parallel to ministries for finance, documentation, engineering, compliance, infrastructure, and HR. Commands arrive through Feishu, Telegram, or Signal channels configured in OpenClaw. The dashboard shows cards by state with heartbeat badges and stop, cancel, and resume controls, plus a department monitor and a memorial archive that turns finished work into five-phase timelines you can copy as Markdown. Nine edict templates add parameter forms and cost estimates, an officials panel ranks agents by merit score while tracking token spend, and per-agent model switching restarts the gateway in about five seconds. Agent thinking, tool calls, and results stream live. A permission matrix limits who may message whom, illegal state jumps are rejected, and every transition lands in an audit log. Worked examples cover competitive analysis, code review, and weekly reports. Dashboard login protects the console, and a systemd unit keeps it running across reboots. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.
Cezar
Close the laptop; the agents keep shipping. Cezar is an open-source control center that runs Claude Code, Codex, OpenCode, Copilot, and eight other coding agents in parallel on your own server, around the clock. Every task gets an isolated git worktree, so a dozen agents edit the same repository at once without colliding, and overflow waits in a queue. You watch each run live: the agent's text, its tool calls, token counts, and reported cost stream into one cockpit covering all your projects. The workspace dashboard separates what needs your answer, what is running, and what finished, and the usage view totals spend and tokens per project so a runaway task shows up in dollars, not surprises. Repeatable playbooks are plain Markdown skill files, chained into short YAML workflows that can mix a different agent per step, add checks and gates, and open pull requests; nothing merges on its own. Hand it a GitHub issue in one click, connect Jira or Linear to launch from your tracker, and schedule recurring maintenance with automations. Flip Autonomous and a run stops asking questions and just finishes; run the same task two or three times and keep the best diff. State lives in plain files, no database, and the responsive interface works from your phone. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.
SandBase Harness
SandBase Harness gives your AI agents a proper home with sessions, tools, sandboxes, and an audit trail instead of a pile of scripts. The console lists agents with their model, status, and version, and an agent page shows its system prompt, attached tools, and skills. Every session keeps a resumable event log, and a transcript view opens any user or agent message as a rendered or raw event. Built-in tools run in a local process, per-session Docker containers, Kubernetes, or a self-hosted worker queue, and permission policies can always allow a tool or always ask a human first. MCP toolsets, credential vaults, and skill packages extend what an agent can reach, with Anthropic's xlsx, pptx, pdf, and docx skills listed next to your own. Memory stores, webhooks, scheduled deployments, and outcome definitions cover longer-running work. Settings include a model provider form, API keys for bearer authentication, and an interactive API reference for sessions, agents, environments, and vaults. A TypeScript SDK and a bundled MCP bridge connect other tools to the same runtime. 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.
EDDI
Deploy autonomous conversational AI agents and coordinate complex multi-agent workflows through EDDI, an open-source orchestration middleware that turns declarative configuration files into secure, production-grade enterprise assistants. Engineering teams connect AI models from twelve different commercial and local providers, using bilateral Model Context Protocol tools to let external desktop clients and coding assistants interact directly with running conversational services. Autonomous agents collaborate through structured group interaction patterns including round table discussions, peer reviews, Delphi consensus rounds, and devil's advocate debates to refine generated solutions before delivery. Declarative workflow extensions automate external REST API calls with dynamic request templating, extracting structured properties into persistent conversation memory and chaining multi-step service interactions without custom script glue. Built-in secret vaults safely manage sensitive API credentials, while the centralized web dashboard provides live conversation inspection, token quota enforcement, and comprehensive audit logging for regulatory compliance. 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.
Manifold
One workspace, a whole roster of AI specialists: Manifold is an agent platform for long-horizon objectives that run through chat, visual workflows, and scheduled tasks. Specialists carry their own model, prompt, and tool allow-list, group into teams, and collaborate across many turns in a chat view that renders HTML visualizations beside text. Models come from OpenAI, Google, Anthropic, or any server that speaks the OpenAI API, such as llama.cpp or vLLM, and image generation reaches OpenAI, Google, or local ComfyUI. The flow editor draws workflows on a canvas, exposes MCP tools as nodes, and turns a saved workflow into a tool that specialists and other flows can call. Pulse schedules specialist tasks at intervals, daily, or once at a chosen time and delivers results to Matrix rooms. Projects isolate agent workspaces to their own folders, load skills from each project's skills directory, and include a file browser with upload, download, and Markdown preview. A Playground versions prompts and datasets and runs experiments that show how a prompt change shifts behavior, while the overview dashboard charts token usage, traces, and logs. Sign-in through OIDC or OAuth2 is optional and adds user and admin roles. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.
Chorus
No agent in Chorus touches your code until it has asked clarifying questions, drafted a plan, and won your approval. Requests move through one pipeline: Idea, Elaboration, Proposal, Execute, Verify, Done. Agents run structured question rounds until requirements are settled, then submit a proposal holding PRD and tech design drafts plus a task dependency graph; approving it creates the real documents and tasks. Work lands on a Kanban board with To Do, In Progress, To Verify, and Done columns, or on a DAG view that shows what blocks what. Live presence indicators reveal which agent is touching which card. Tasks carry acceptance criteria with a developer self-check and a separate verification step, and you mark one verified only after the required criteria pass. Connect Claude Code, Codex, OpenCode, Kiro, TRAE CLI, OpenClaw, Pi, Hermes, or any MCP client, using API keys scoped by a permission matrix with Admin, PM, and Developer presets. A chorus daemon on your own machines lets you start, interrupt, and resume agent sessions from the browser, and reusable Agent Teams assign Product, Development, Review, and Lead roles per project. Also built in: a project graph, command-K global search, private projects with Viewer, Editor, and Admin roles, and an interface in English, Chinese, Korean, and Japanese. 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.
Clodfarm
While you sleep, a crew of Claude Code agents in a single Docker container keeps building: you plant a mission, the farm splits it into sub-agents on their own branches, and only work that passes your tests ever lands in the repo. Open the farm in your browser and every agent is a pixel Claude walking its plot, tending a task at a terminal or napping when the budget says stop, with the quest log, mission board, and a hatchery for new agents a click away. Each account is paced against its real five hour and weekly usage, measured live, always leaving you twenty percent and never paying overage, and teammates' agents message each other and hand jobs to whichever account still has room. Talk to any agent from the Claude app on your phone, post quests, set schedules like weekday issue triage at nine, and watch live dashboards the agents maintain themselves. Whiteboards let you sketch while an agent draws back full Mermaid architectures for the whole farm to see, the farm's logged-in browser does website work as you, and Slack mentions, Blender scenes, and other models from GPT to a local Ollama all plug into the same queue. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.