The full InfiniBot agent OS on your own private cloud machine. Sign in, bring your AI keys, and a fleet of agents — Claude Code, Codex, OpenCode, Gemini — goes to work. Nothing to install.
Private machine per workspace · Wakes in seconds · Your own keys
Flat from $19/mo · AI usage bills to your own provider · Free to self-host
Nothing to install
The whole OS in a browser tab — desktop, tablet or phone.
Your own machine
An isolated cloud machine per workspace. Never a shared seat.
Wakes in seconds
Sleeps when idle on Starter, or stays on 24/7 from Pro up.
your-workspace / agent meshresuminglive
example run
machineresumed from sleepready
claudeaudit the auth flowdone
codexfix the failing testsdone
opencodescreenshot every pagedone
telegramreport sent to yousent
Works with
Claude CodeCodexOpenCodeGemini CLIAnthropicOpenAIOpenRouterGroqGoogle GeminiDeepSeekTelegramWhatsAppSlackMCPClaude CodeCodexOpenCodeGemini CLIAnthropicOpenAIOpenRouterGroqGoogle GeminiDeepSeekTelegramWhatsAppSlackMCP
Live demo
One person. A fleet of agents. In a browser tab.
This is the real InfiniBot agent mesh: coder agents fanning out across a repo audit, each walking to its room on the map as it works. In the cloud, the same mesh runs on your private machine — open a tab and it's there.
Local agents are powerful right up until the lid closes. Then the fleet stops, the bots go quiet and tonight's work waits for morning.
01
Lid closed, work stopped
Local agents live and die with your laptop. Close it, lose Wi-Fi or let it sleep and the fleet stops — mid-run, with the work half done.
02
Setup is its own project
Runtimes, CLIs, ports, tunnels and a gateway to babysit before the first agent runs. Then do it again on the next machine.
03
Keys in every .env
API keys copied into config files on every device you own. Nothing encrypted, nothing in one place, no idea which key an agent used.
04
No front door
Reaching your agents from a phone means VPNs and port forwarding. Most people just don't — so the work waits for the desk.
The fixInfiniBot Cloud moves the whole OS onto a private machine that is always one tab away — from any device.
02Your own machine
One workspace. One private machine.
Every workspace gets its own isolated cloud machine with its own storage and gateway — not a seat on a shared chatbot. It sleeps when you are idle and wakes in seconds when you come back.
You
Browser
Telegram
Voice
Front door
Sign-inverifies it's you
Routerfinds your machine
Your machine
workspace machinelive
gatewayagentsmemorychannels
> sign-in verified
> machine resuming…
> gateway healthy
> workspace live
Isolated by default
Its own processes, storage and gateway. Your agents, files and memory never share a runtime with anyone else.
Wake or always-on
Starter wakes when you open the app and sleeps when you leave. Pro, Business and Scale stay on 24/7, so agents keep working while you're away.
The same OS as self-host
Not a lite edition. It's the full InfiniBot you can run yourself, with state kept on the machine's own volume.
03Swappable stack
Swap the agent. Swap the model. Same fleet.
Install Claude Code, Codex, OpenCode or Gemini CLI from Setup › Coders in one click, sign in, and each becomes a backend the orchestrator can hand work to.
Layer ACoding agents — four, swappable per task
ClaudeCode
Anthropic
Deepest reasoning — long-context planning and careful, reviewable edits.
claude_coder
Codex
OpenAI
Fast coding bursts — rapid, high-throughput implementation passes.
codex_coder
OpenCode
Any provider
Open and provider-agnostic — route it to whichever model you bring.
opencode_coder
Gemini CLI
Google
Google's open-source terminal agent, as another backend for sub-agents.
gemini
Setup › Coders
One click to install. One sign-in to use.
Coders install onto your cloud machine, not your laptop. Sign in once and every sub-agent that needs that backend can use it.
Install, update and sign in from one page
Pick the backend per task or let the orchestrator choose
17 model providers. Each key is checked with the provider when you add it, then encrypted.
04Agent fleet
A fleet that runs itself
Hand the orchestrator a goal. It splits the work, spawns sub-agents with non-overlapping scopes, and checks each result before calling it done.
ORorchestratorClaude Coderunning
› splitting the launch checklist into scopes
UIui-architectClaude Coderunning
› restyling the pricing cards
CXcodex-refactorCodexrunning
› fixing the failing auth tests
QAqa-inspectorOpenCoderunning
› screenshotting every page at 390px
RSdeep-researcherClaude Coderunning
› comparing three payment providers
GMgemini-docsGemini CLIrunning
› drafting the API reference
CRcrm-managerOpenCoderunning
› logging new leads to the pipeline
TGtelegram-relayChannelrunning
› sending you the daily report
Illustration of a run — your agents, tasks and counts will differ. Parallel agents scale with your plan.
Parallel by default
Research, build, audit and ship run side by side instead of waiting in one chat thread.
Verified, not just claimed
Runs finish with evidence — commands, tests, screenshots — and get reviewed before they count as done.
Steer mid-run
Message any running agent, stop it, or respawn it with its progress so far. You stay in charge.
05Live agent mesh
Watch every agent on the map
Agents appear on a live isometric map, walk to the room that matches their work and report back when they finish. Click one to open its chat panel right on top of the map.
your-workspace / agent meshlive
chat panel / on the map
Every agent, room and status at a glance
Chat panels for any agent, anywhere in the app
Filters for activity, burn and failures
Your last 10 finished runs stay on the map, so you can see what happened while you were away
Captured from a fresh InfiniBot Cloud workspace.
06Tools & integrations
One request. Every tool.
Say what you want done. The orchestrator picks the right agents and fans out to coders, native tools and MCP servers — in parallel, plus the agent browser and your channels.
Add any MCP server and its tools become available to every agent you allow.
07Voice orchestrator
Delegate by voice. Stay in flow.
Talk to it like a colleague. It turns a spoken sentence into routed, scoped work — and tells you when it's done.
Hands-free delegation
No tab, no form, no typing. Push-to-talk and the orchestrator takes it from there.
Parallel sub-agents
Heavy work fans out to background agents at once, each on its own scope.
Spoken progress
The line stays open while agents run, and each result is read back as it lands.
Voice runs on Robovoice, linked to your workspace.
You say
“Ship the landing tweaks and run the tests.”
orchestrator
Claude Codeedit the landing page
Codexrun the test suite
OpenCodelint + typecheck
Claude Codebuild the bundle
It reports back
“Done. Landing shipped and the tests are green.”
08Everywhere you are
Your agents, in your pocket
Connect a Telegram bot, WhatsApp or Slack and talk to your workspace from anywhere. On always-on plans they answer at 3 a.m.; on Starter they wake with you.
Approvals on Telegram
The app on your phone
Telegram
Chat with the orchestrator, get reports and approve actions with one tap.
WhatsApp
Link your number and message your workspace like any other contact.
Slack
Bring agents into the channels your team already works in.
Owner-only approvals
Risky actions stop and ask you first
Deletes, spends and other gated tool calls wait in the approvals inbox in the app — and the same card lands on your Telegram. Only the workspace owner can approve.
approvals / inbox
09It knows itself
InfiniBot shows you around
Ask where something is and it doesn't describe a menu — it opens the page and highlights the exact control, step by step. It knows its own pages, settings and plugins, so setup becomes a conversation.
Ask“How do I add my Anthropic key?”
It navigatesOpens Keys & Usage for you.
It highlightsSpotlights Add key and walks you through it.
keys / guidedlive
solatrix / settings
10What's inside the OS
The full agent platform that does the work — not just chat
Everything below ships in every cloud plan. Each screen is a real capture of InfiniBot.
F-01Projects & kanban
Projects your agents actually update
Projects, tasks and a kanban board live inside the OS. Agents file their own cards, move them as they work and leave notes you can read back.
Every card can link to the agent doing the work
Rich task pages with content blocks
Kanban, list and board views
infinibot / projects
F-02Knowledge graph
One graph of everything you know
Projects, clients, notes, tasks and research become one navigable graph, so agents and you can follow the thread from any node.
Auto-linked from your real data
Filter by type, zoom into clusters
The same graph agents query
infinibot / graph
F-03Shared memory
Memory every agent shares
What one agent learns, the next one knows. Memory is shared across the whole fleet and lives on your machine's own volume.
Per-agent knowledge plus a shared store
Knowledge sources you can inspect
Persists across sleep and wake
infinibot / memory
F-04Notes & documents
Documents agents read and write
A rich document editor with categories, tags and project links. Agents draft specs, reports and plans straight into it.
Rich editor with slash commands
Link docs to projects and tasks
Search across everything
infinibot / docs
F-05Workflows
Workflows that run on schedule
Save a multi-step process once and run it on demand or on a schedule. On always-on plans they fire while you sleep.
Inputs, steps and verification
Cron schedules on Pro and up
Run from chat, voice or Telegram
infinibot / workflows
F-06Activity
See every spawn, call and result
A live feed of sub-agents spawning, tools being called and runs finishing, with the brief and transcript one click away.
Sub-agent history with outcomes
Tool usage per agent
Jump straight into any run
infinibot / activity
F-07Tools & MCP
Tools that plug right in
Kanban, notes, memory, research, workflows, browser and more are native tools. Add MCP servers and skills on top.
Native tool registry
MCP servers and a skill library
Per-agent tool access
infinibot / tools
F-08Agents
An org chart of AI agents
Give agents roles, models and tools, arrange them in a hierarchy and let the lead agent triage and delegate to the rest.
Roles, models and tools per agent
Lead agent delegates and reviews
Chat with any agent directly
infinibot / agents
11Plugins
Add whole apps with one switch
The plugin catalog adds first-party apps with their own pages, agents and data. Turn on what you need; everything else stays off.
SX
Solatrix
Solana market intelligence
Holder bubble maps, convergence boards, AI analysis and alerts — with agent access you control, from read-only to full, and a kill switch.
solatrix / bubble map
BA
Business & Agency
Run an agency from the OS
Executive view, daily operations, an AI workforce with roles and shifts, approvals, reviews and an agency topology — driven by an operating loop you authorise.
command / today
Setup › Plugins
A catalog, not a rebuild
Browse, enable and disable plugins from the catalog. Each one declares its own pages, tools and permissions, so you know what it can do before you switch it on.
setup / plugins
Ready when you are
Seen the tour? Open your workspace.
Pick a plan and your private machine is provisioned for you. Bring a key and put the fleet to work.
Every plan includes the full OS: coders, mesh map, chat, memory, plugins, MCP and terminals, plus voice, channels, scheduled jobs and the agent browser. BYOK — AI usage is billed by your own provider.
InfiniBot Cloud is the full InfiniBot agent operating system running on your own private cloud machine. You get parallel coding agents (Claude Code, Codex, OpenCode and Gemini CLI), a live agent mesh map, chat, projects, memory and a knowledge graph, and plugins — in the browser, with nothing to install. Voice and Telegram, WhatsApp and Slack channels are included too.
Do I need to install anything?
No. Sign in at the app and your workspace runs on a cloud machine we host for you. Coding agents like Claude Code are installed onto that machine from Setup › Coders with one click.
Is my workspace shared with other customers?
No. Every workspace runs on its own isolated cloud machine with its own storage and gateway. Nothing is pooled with other workspaces.
How fast does my machine wake up?
A sleeping machine resumes in seconds when you open the app. On Starter it sleeps when you are idle; on Pro, Business and Scale it stays on 24/7, so there is nothing to wake.
Do I need my own AI keys?
Yes. InfiniBot is bring-your-own-key. Add keys for providers like Anthropic, OpenAI, Google Gemini, OpenRouter or Groq and they bill you directly at their price. InfiniBot charges only the flat plan price and never marks up tokens.
How are my API keys stored?
Each key is checked with its provider when you add it, then encrypted at rest with AES-256-GCM and only ever shown masked. Model calls are made on your key from your workspace.
Who can approve risky actions?
Only the workspace owner. Actions that need a sign-off wait in the approvals inbox in the app, where the owner approves or denies them. The same approval card is sent to your Telegram so you can decide from your phone.
What is the difference between Cloud and self-host?
Self-host is free under AGPL-3.0 and runs on your own hardware, including local models. Cloud runs the same OS on a private machine we manage, reachable from any browser, with always-on plans that keep your agents running while your computer is off.
How many agents can run at once?
Starter: 4 parallel agents · 1 coding at once; Pro: 8 parallel agents · 2 coding at once; Business: 16 parallel agents · 4 coding at once; Scale: 24 parallel agents · 12 coding at once. Coding agents like Claude Code are full CLI processes and use far more memory than a native agent, so they have their own, lower cap. More parallel agents means a bigger plan.
Can I cancel or change plans?
Yes. Plans are monthly and billed through Whop. Upgrade, downgrade or cancel at any time.
Get started
Put a fleet to work tonight
Sign in, bring your keys, and your private agent OS is ready in minutes. Nothing to install.
