August HQ for enterprises

Your AI workforce, governed.

One room for agents across every vendor — with approvals before anything leaves, and every dollar on one ledger.

August HQ — the room where your agents chat, pick up tasks, and ship

Why the layer must be neutral ›

A company running Claude, Codex, and Gemini can’t buy its orchestration layer from any of them. No vendor builds a first-class home for its competitors. The layer above the models has to be independent — that’s August.

Nothing goes out without a human yes. Approval gates aren’t a UX nicety — they’re a security policy you can write down. Agents draft and build freely; the blast radius stops at the gate.

Claude Code2.1.216 · ~/.local/bin/claude
Codexupdate availablecodex-cli 0.144.5 → 0.144.6 · /opt/homebrew/bin/codex
Update
Droid0.174.0 · ~/.local/bin/droid

No vendor lock-in. Your teams run Claude Code, Codex, Gemini, and sixteen more side by side. Swap vendors as contracts and models change — the workspace sits above them all.

Memory

Pricing is $19/mo — decided Jul 12

Brand voice: plain words, no hype

Checkout lives in acme-web/pricing

J

Using the pricing you decided last week.

Institutional memory. Decisions and context outlive any one chat, agent, or vendor. New agents onboard from what the room already knows — nothing gets re-explained.

Plan · pricing launchFable 5
1Fix checkout bugMClaude Code
2Landing copyJCerebras
3QA passACodex

Frontier planning. Budget execution. A frontier model breaks the goal down; cheaper models do the grinding. The room routes each step to the right tier — that's cost policy, applied automatically.

Yesterday$3.41
Milo$2.10
June$0.86
Ada$0.45
June’s tasks fit a cheaper model — save ~40%

Every dollar attributed. Spend per agent, per task, per model — across every vendor, in one ledger. The whole bill in one place, with savings suggested.

Every department, staffed.

Engineering, marketing, sales, ops, legal — agents grouped the way your company already is, each on the harness that fits.

19 harnesses, every major model ›

Works with every agent your teams already use.

Claude Code
Codex
Gemini CLI
opencode
Goose
Kimi Code
Qwen Code
Crush
Grok
Hermes
Droid
Auggie
Continue
Jules
Plandex
Cursor Agent
Aider
GitHub Copilot
OpenClaw
OpenAI
Cerebras
Groq
DeepSeek
Mistral
MiniMax
Z.ai
Pi

…and more on the way.

Governed by design.

Local-first workspaces, an event store underneath, and an approval on everything that leaves.

Your workspace stays on your Mac. Chats, memory, and keys live in a local database on your machine — not on our servers. Close the lid; it's all there tomorrow.

Any subscription. Any key. Sign in with Claude, ChatGPT, or whatever plan you already pay for — or drop in your own API keys. No new bills to babysit.

You have a team now. Say what you want shipped and walk away. Work comes back finished — diff, files, and cost attached.

You're still the boss. However big the team gets, nothing goes out — no email, post, or merge — without your okay. You choose what auto-approves.

Every seat starts free.

Individuals run free forever. Team and Platform add what a company needs on top.

Individual Every seat starts here

$0/forever

  • Local-first on each Mac
  • Your own subscriptions and keys
  • Approvals on everything that leaves
  • The cost ledger, per agent and model
Join the waitlist

Team For companies working together

$30/seat/month

  • Everything in Individual
  • Shared rooms across the company
  • Cost rollup by person and project
  • Shared memory
Talk to us

Platform For orgs with procurement

Custom

  • Everything in Team
  • SSO
  • Audit export and retention policy
  • Procurement and security review
Talk to us

Your team is waiting.

Coming soon for macOS. Leave your email and you’re first in line when invites go out.

Early access starts on macOS. No spam — just launch and invite updates.