Agents are fast now. You're the bottleneck.
You run an agent in every window. Sessions sit idle while you answer another one, and you repeat the same context in every tool. Your attention is the only limit left.
Should the free tier stay on the pricing page?
Plans done. Before I continue: free tier or trial?
Tests pass. Open the PR?
Running the test suite…
Draft ready. Publish today or Friday?
Migrating fixtures…
Should the free tier stay on the pricing page?
Illustrative example
Every agent wants you at once.
Agent turns are slow, so you start more work while you wait. Then each agent finishes at a different moment and asks for you.
Idle sessions wait on you.
You answer one session at a time. The other sessions sit waiting, and some get forgotten for hours.
You repeat yourself in every tool.
Grok never heard what you told Claude. You explain the same context again, and set up the same connectors again.
Run every agent from one conversation.
Hand off the work, and agents report back to you. Ask what's next, and every decision that needs you comes back numbered, with the context to decide. Your answers become rules every agent follows.
Start with the free primer. The kit comes with the guide.
api#90 · review passed · ruling 3
Sent an agent onto api#88: flaky checkout tests.
Needs you (1)
1. Merge the webhook retries PR?
An independent review passed. Ruling 3 says agents don't merge to main without you.
api#90 · review passed · ruling 3
A: Merge it (recommended) · B: Hold for your own review
In flight
api#88 Flaky checkout tests · codex · working
✓ Recorded 1: merge the webhook retries PR. Merged.
Illustrative example
Hand off the work. Then ask what's next.
Keep the problems you want to solve yourself. Hand off the rest.
-
01 · HAND OFF
"Fix the flaky checkout tests and open a PR."
Your session sends an agent onto the issue and tracks the run.
-
02 · WORK RUNS
Agents work in the background.
Every run shows as in flight, returned, in review or landed. You move on.
-
03 · REPORTS BACK
A fresh agent checks the return.
Anything that needs you comes back as a numbered question, with the context to decide and links to the issue, PR or earlier ruling.
-
04 · YOU ANSWER
"1 B."
Your answer is recorded word for word before anything acts on it. Answer several at once, or in your own words when the question is off base.
Then ask what's next? and go again.
Watch one morning play out two ways.
Overnight, your agents finished the pricing work. Two decisions need you: the free tier and the changelog. Same agents, same morning, before Command Tempo and after.
A session for every agent, and each session waits on you.
Should the free tier stay on the pricing page?
Done with the Stripe plans. Before I continue: should the free tier stay?
Draft ready. Publish today or hold until Friday?
Churn is up 4%. Want the breakdown by plan?
Tests pass. Want me to open the PR?
Three drafts below. Which tone fits?
Self-hosted Postgres or SQLite for now?
All five emails drafted. Review the copy?
Numbers pulled. Add the churn note?
- Hunt through sessions for the agent that needs you.
- Find the site agent idle two hours later.
- Explain yesterday's answer again in another session.
One conversation, both decisions numbered, one reply.
Ruling 2 says keep the free tier until launch. Launch is Monday.
site#212 · billing#48 · ruling 2
Needs you (2)
1. Keep the free tier on the pricing page?
The site and billing agents both need this answer. Ruling 2 says keep the free tier until launch, and launch is Monday.
site#212 · billing#48 · ruling 2
A: Keep it · B: Drop it, keep the trial (recommended)
2. Publish the changelog today?
A: Yes (recommended) · B: Hold until Friday
In flight
This week's figures · working
✓ Recorded 1 as ruling 5: drop the free tier, keep the trial. Replaces ruling 2.
✓ Recorded 2: the changelog goes out today.
Site and billing agents back at work.
- Ask what's next and see both decisions, numbered.
- Answer both with "1 B, 2 A".
- Give the free tier answer once. Every agent follows the new ruling.
Set the tempo. Work live, or check in when it suits you.
Work beside your agents.
Take one question at a time when you're stretched, or clear a batch in one sitting. Agents keep working on everything your answer doesn't block.
Step away while the work runs.
Agents work on a schedule while you sleep or build something else. Anything they can't finish alone waits in one queue: decisions, reviews, sign-offs, merges, posts. Clear the queue a few times a day. You pick the times when you set up the schedules.
Every answer becomes a rule.
Stop answering the same question twice. Claude Code and Codex read the rules in force at the start of every session.
- Change your mind any time. The new ruling replaces the old one, with the old one quoted.
- Search what you decided and get your own words back.
- Store every ruling in plain files in your own git folder.
Ruling 5 · Free tier
"Drop it, keep the trial."
Plain files and two skills, in your own repo.
One command adds these to your project. Claude Code and Codex pick them up the next time a session starts. Your agents run on your own Claude or Codex plan, and the kit never runs a model for you.
rules-in-force.md- One line per standing ruling. Every session reads the list at startup, through
CLAUDE.mdfor Claude Code andAGENTS.mdfor Codex. decisions.md- Every ruling in full: your answer word for word, the date, where you answered and who decided.
decision-log.md- Calls your agents made on their own, with their reasoning, for you to review.
MODE- One word.
run: agents act within your rulings.shadow: any push, post, email or purchase outside the repo comes to you as a question first.halt: everything stops. .claude/skills/- The instructions agents follow to ask you, record your answer and show what's next. Codex reads the same files.
tools/decisions/- A Python script and an MCP server to ask, record, search and check. Standard library only.
- A pre-commit check
- Blocks a commit when a ruling is missing from the rules list or a ruling has no source.
- Schedule templates
- A morning gather, an evening digest and a health check, for cron, launchd or Claude routines. Off until you switch them on.
Start with the free primer. Go further when you're ready.
Prices in Australian dollars.
Try the method in your next session.
A short read on running every agent from one conversation: the loop, numbered decisions and the decision log.
Pay once
- The whole method, step by step: the loop, the decision log, roles and schedules.
- Write your run book and your first rulings with the templates.
- Use the method with plain files and the agents you already run.
Pay once
- Everything in the guide.
- The Command Tempo kit, installed into your project with one command.
- Rulings, what's next, dispatch and review, the setup interview and schedules.
- Read and change the code under the Functional Source License.
Monthly
- Your rulings, context and run book in one place, for every agent you run.
- One sign-in for every tool.
- Agents keep working while your computer is off. Answer from your phone.
- Teams: you own your calls and share the ones you choose.
No GST charged
- We install the kit and connect your tools with you.
- We write your run book and record your first rulings with you.
- You never touch git.
Read before you sign up.
What does Command Tempo add to my project?
The files listed under What you install: three Markdown files, a MODE file, two skills and a small Python tool. Nothing runs in the background unless you switch on the schedules.
How do agents know my rulings?
Each session starts by reading rules-in-force.md, one line per standing ruling. The full record in decisions.md is searched when an agent needs it, never read in full. A ruling outranks every other instruction except MODE.
How much context does it use?
The rules list is capped at 1,500 words and read once per session. A search returns only the rulings that match.
How do the numbers work?
In a session, decisions show as 1, 2, 3, and "2 A" answers number 2. A decision still open when the session ends moves to a lasting number: its GitHub issue number when your project is on GitHub. You never see internal IDs.
What if a question is wrong?
Answer in your own words: "2: neither, ask the client first." Your words go on the record as the answer, and the agent works from them.
Which agents work with Command Tempo?
Claude Code and Codex run the conversation today. Any agent with access to the folder or the repo reads your rulings, such as Grok with a GitHub connector or a cloud agent with repo access.
Can I answer away from my computer?
Yes. Answer on the decision's GitHub issue, or through your own Slack connector and a scheduled task. The hosted platform will add answers from your phone with nothing to set up.
Where does my data live?
In plain files in your own git folder. The kit runs no server of its own.
Do you set up my connectors?
We ship guides for GitHub, Linear, Granola, Slack and email, through MCP or Composio. You add each connector in your own agent. The done-for-you setup through Full Throttle CEO connects your tools with you.
What will the hosted platform add?
Your rulings, context and run book kept for you, so every agent on every device reads the same thing. One sign-in for every tool through Composio, which holds your logins and passes each tool to your agents. Agents in cloud runners that keep working while your computer is off, and answers from your phone. Teams that share by role and by class: metrics, release dates, product behaviour and technical scope. See the hosted platform.
Is the kit free?
The primer is free. The kit comes with the guide, as Guide + repo. The code is source-available: you can read it before you buy, and run and change it for your own use.
What does the Functional Source License allow?
You can read, run and change the kit's code for your own use. Each version becomes open source two years after release.
Can I stop every agent at once?
Write halt in the MODE file. No agent edits a file or sends a message until you change it back.
Get the free primer.
A short read on running every agent from one conversation. Download it as soon as you sign up, and we'll tell you when the guide opens.