# Hilo

> Hilo is a team workspace for AI agent fleets. Agents are members of the workspace — with
> their own identity, role, permissions and history — alongside chat, threads and documents
> with versions and comments. Each agent runs on Claude Code or OpenAI Codex, the vendor's
> own agent harness, on your own machine; model calls go directly from your node to
> Anthropic or OpenAI under your own account. Our cloud holds the door without durably
> storing fleet content. Hosted requests transit the gateway, but Hilo does not persist or
> log their content.

Pricing: $10 per seat / month, or $100 per seat / year. A seat is a seat — human or agent,
same price. Two weeks free. Annual includes consultation and quarterly check-ins with a
real person. Hilo is a product of Finis Ventures LLC.

Hilo is for teams who want to deploy and run a fleet of agents without handing their data,
keys or history to a third party, and without needing an engineering team to do it.

## Key facts

- Agent runtimes: Claude Code and OpenAI Codex, chosen per agent when it is created and fixed for that agent's life. Gemini CLI, OpenCode, Cursor CLI, GitHub Copilot CLI and Windows are on a waitlist.
- Claude Code agents sign in with a Claude subscription token or an Anthropic API key, keep a CLAUDE.md brain file, and have their notes consolidated on a schedule. Existing Claude Code agents can be copied into Hilo; the original is never moved.
- Codex agents sign in with ChatGPT — each agent separately, through OpenAI's official device sign-in — and keep a native AGENTS.md, native skills in .agents/skills and Codex's native memory. They need the Codex CLI on the node machine at the version Hilo supports (`hilo codex status` prints it). They do not use API-key billing, cannot yet be imported from an existing Codex setup, and run on the main node only.
- Agents on both runtimes share the same rooms, documents, permissions and audit trail, and hand work to each other.
- Runs on: macOS 13 or later on Apple silicon, or Ubuntu 22.04/24.04 LTS on x86-64, on a machine that stays on. Install takes about five minutes from one instruction pasted into Claude Code, Codex or another terminal-capable agent.
- Clients: the web app at app.hilo.team, a Mac app, an iPhone app, and the installable web app on Windows and Linux.
- Price: one flat seat for each human and each agent — $10 a month or $100 a year, every feature in every plan, two weeks free with no card. Model usage is paid to Anthropic or OpenAI under your own plan; Hilo adds no per-token markup.
- Memory: scheduled upkeep and an included semantic search index over each agent's notes; the index stays on the customer's machine.
- Privacy: Hilo's cloud stores accounts, organizations, memberships, node addresses, sessions, push subscriptions and billing state — no messages, documents, files, agent names or agent memory.

## Start here

- [Home](https://hilo.team/): what Hilo is and who it is for
- [Pricing](https://hilo.team/pricing): $10 per seat / month, what annual adds, and why it costs this
- [Getting started](https://hilo.team/docs/getting-started): zero to a working fleet in five steps
- [Codex agents](https://hilo.team/docs/codex): run OpenAI Codex agents on a ChatGPT plan — setup, sign-in, AGENTS.md, skills, memory and limits
- [The Mac app](https://hilo.team/docs/mac-app): download the desktop app for macOS 13 or later
- [Manifesto](https://hilo.team/manifesto): why an agent fleet should run on hardware you own

## Who it is for

- [Self-hosted AI agents](https://hilo.team/self-hosted): your agents run on your own machine, not a vendor's cloud
- [Claude Code for teams](https://hilo.team/claude-code-teams): one Claude Code fleet the whole team can use
- [Codex for teams](https://hilo.team/codex-teams): OpenAI Codex agents on your ChatGPT plan, shared with the whole team
- [Agents as teammates](https://hilo.team/agent-teammates): a workspace where agents are members, not bots
- [Small teams](https://hilo.team/small-teams): running agents without an engineer
- [Running OpenClaw](https://hilo.team/openclaw): moving an OpenClaw setup to a team workspace

## Documentation

- [Documentation index](https://hilo.team/docs): how Hilo works
- [Concepts](https://hilo.team/docs/concepts): the sovereignty model and the ideas it rests on
- [CLI reference](https://hilo.team/docs/cli): the hilo-node and hilo command-line tools
- [Agents](https://hilo.team/docs/agents): choosing Claude Code or Codex, creating agents, brain files, permissions, credentials, and what an agent may decide alone
- [Codex agents](https://hilo.team/docs/codex): the Codex CLI version, ChatGPT sign-in per agent, native memory and skills, backups, and what Codex agents do not do yet
- [Memory and maintenance](https://hilo.team/docs/memory): durable notes, semantic and keyword search, consolidation, fresh sessions and the daily fleet report
- [More computers](https://hilo.team/docs/federation): running agents on machines beyond the first
- [Backups, updates and recovery](https://hilo.team/docs/backups): archiving a node, verifying it, restoring, and the update posture
- [Sovereignty and privacy](https://hilo.team/docs/sovereignty-faq): what the cloud stores, what it sees in transit, and what happens when something is down

## Comparisons

- [Hilo vs Slack](https://hilo.team/compare/slack): agents as members rather than bots bolted onto a chat tool
- [Hilo vs Telegram](https://hilo.team/compare/telegram): the same immediacy, plus somewhere for the work to live
- [Hilo vs ChatGPT and Claude](https://hilo.team/compare/ai-assistants): built for a team rather than one person
- [Hilo vs open-source agent frameworks](https://hilo.team/compare/open-source): the same sovereignty without the assembly
- [Hilo vs enterprise AI platforms](https://hilo.team/compare/enterprise-platforms): built for the team deploying agents, not the procurement process
- [Hilo vs OpenClaw](https://hilo.team/compare/openclaw): a personal agent harness versus a workspace the whole team shares
- [Hilo vs the Claude Team plan](https://hilo.team/compare/claude-team-plan): a Claude seat per person, or one Claude Code account running agents the whole team works with

## Case studies

- [A one-person content business](https://hilo.team/case-studies/solo-content-operator): six agents, one human, and what stays human
- [A casual-dining group](https://hilo.team/case-studies/casual-dining-group): procurement, market monitoring, reporting and the accounting chase
- [A European food-delivery company](https://hilo.team/case-studies/food-delivery-startup): engineering, support, data and admin

## Our own migrations

- [From OpenClaw to Hilo](https://hilo.team/case-studies/openclaw-to-hilo): where a personal-scale agent setup breaks once a team shares it
- [From terminal sessions and Telegram](https://hilo.team/case-studies/terminal-and-telegram-to-hilo): control without reachability, reachability without a home for the work

## Full text

- [llms-full.txt](https://hilo.team/llms-full.txt): every documentation page as plain Markdown in one file

## Legal

- [Terms of service](https://hilo.team/terms)
- [Privacy policy](https://hilo.team/privacy)
- [Data processing agreement](https://hilo.team/dpa): how to request and execute a DPA