FAQ

Not for real work yet. Version 0.0.3 is an early preview: anything may change between releases, including commands, config formats and where files live, with no migration path. Try it, but don’t rely on it.

Wheems is free and open source, under the Apache 2.0 licence. Agents run on the engine subscriptions you already pay for: a Claude subscription for Claude Code, a ChatGPT subscription for Codex, or OpenRouter, a provider’s key or free models for opencode. Slack’s free plan works. Search by meaning with Ollama runs on your Mac for free.

Herdr core runs each agent in a terminal that stays alive, which Wheems can start, send to, read and wait on, and which you can attach to at any time to watch or take over. Agents run in their own CLIs, unchanged, so you use the subscriptions you already have rather than paying per token.

No. Wheems is for business work generally: an inbox, reviews, leads, bookkeeping, a website. Agents happen to run in coding CLIs because those are the most capable agents you can run on a subscription.

It runs on macOS today. The service needs launchd. Linux, and running on an always-on server, is planned for milestone 10. Windows isn’t on the roadmap.

No. Until milestone 10, the service runs on your Mac, so agents answer only while it’s awake. Slack retries delivery only briefly, so a message sent while the service is offline usually gets no answer: send it again when it’s back.

Not yet. Slack is the only channel today. Telegram and email are planned for milestone 10, Teams for after version one. Each is a channel driver, so others can be added without changing the core.

Your Slack messages go through Slack. Agents run on their model providers (Anthropic for Claude Code, OpenAI for Codex, whichever provider opencode is set up with), so those providers see each conversation and what agents load or read, including about-you.md and the brief. Search by meaning with Ollama is local: nothing you index is sent anywhere.

In ~/.config/wheems/secrets.json, readable only by your user, outside your install folder and out of git. You type them into wheems setup; they never pass through an agent and are never in what agents are given. But agents run as your user, so an agent that went looking could read the file: it isn’t locked against them until the vault (milestone 9) and permission rules (milestone 6). See Connecting Slack.

Only people whose Slack member IDs are in allowed_users. Messages from anyone else are ignored and logged. See Allowing people.

Today, yes, if they went looking: workspaces are separate folders, not walls. Enforced isolation arrives in milestone 6. Until then, don’t rely on workspaces to keep secrets from each other.

No. One Slack app serves every Wheems workspace: one gets its direct messages, the others a channel each. See One Slack app for every workspace.

Yes. Each agent names its engine, so one workspace’s chief of staff can run on Claude Code and another’s on Codex. See Engines and accounts.

Yes. Pages are Markdown in git. The filer never overwrites your edits, and search picks them up within ten minutes.

hc ls lists the sessions; hc attach <session> shows one live, and you can type into it. Press Ctrl+\ to detach. Wheems’s notices in Slack include the right hc attach command.

wheems service uninstall
mv ~/wheems ~/wheems-old
wheems setup

Your Slack tokens stay in ~/.config/wheems/secrets.json, so setup offers to keep using the same app.

github.com/wheems-dev/wheems. Issues and ideas are welcome there.

Yes: Wheems was formerly called Opinr.