Skip to main content
Closed beta · 50 founder agencies

Your best prompts shouldn't live in one person's head.

Loops is the team-shared layer on top of Claude and ChatGPT. Describe a workflow in English. Loops tests it, versions it, tunes it to your client's brand, and ships it to every platform your team uses.

£25/seat lifetime for the first 50 agencies · 13 claimed
loops · weekly-client-report.skill.md
--- name: weekly-client-report version: 1.0.3 team: Kite Studio · 8 seats brand: ACME · tone: crisp, factual author: Maya Okafor --- // Take a client's weekly status update and turn it // into a formatted report in their brand voice. // Passed 12/12 reference evals · updated 2h ago  
Built on the open Agent Skills standard
Claude Desktop Claude Code ChatGPT OpenAI Codex

Every agency has the same dead Notion page.

It's called "Prompts." Nobody uses it. Your best prompts live in one strategist's head — the one who got the ACME account to sound like ACME. When she's on holiday, the rest of the team reverts to generic output. When she leaves, six months of client-tuned work walks out with her.

You don't have a prompt problem. You have a shared-asset problem.

Shared Notion pages are dead. Custom GPTs lock you to one platform. Your agency's best AI work shouldn't live in someone's chat history — it should live as a team asset: tested, versioned, brand-tuned, and available across whichever tool a teammate opens on a given Tuesday.

Describe it. Test it. Ship it everywhere.

Three steps, end to end. No boilerplate. No deploy scripts. The authored Loop runs identically across every target platform your team uses.

01 · Describe

Describe the workflow in English

"Turn a client status update into a formatted report in their brand voice." Loops builds the SKILL.md for you.

02 · Test

Test against your good examples

Paste three outputs you already liked. Loops audits the skill against them, flags drift, tunes until it matches your bar.

03 · Deploy

One click — every platform

Deploys to Claude Desktop, Claude Code, ChatGPT, and Codex via the Agent Skills open standard. Every teammate runs the same Loop.

Ships to every platform your team already uses.

Loops builds on the open Agent Skills standard (December 2025). One authored Loop runs identically across Anthropic's and OpenAI's products — no rewriting prompts for different tools.

C

Claude Desktop

Auto-installed for every teammate. Invoked by name in any chat.

</>

Claude Code

Available as a slash command in any repo your team opens.

G

ChatGPT

Loaded as a Custom GPT skill — same output, same voice.

CX

OpenAI Codex

Via the December 2025 cross-platform skills standard.

Nothing else does all four.

Most agencies have already tried the obvious options. Here's why they keep breaking — and what Loops does differently.

Shared Notion Custom GPTs Team Slack channel Loops
Team-shared, not tribal ~In theory Yes No Yes
Version-controlled Page history only No No Git-style
Tested against good outputs No No No Built in
Per-client brand profiles Manual ~Per-GPT only No First-class
Cross-platform (Claude + ChatGPT + Codex) No ChatGPT only No All four
Works when your best person leaves No ~If they published it No Yes

Priced for real teams, not solo hobbyists.

Solo is free forever. Team and Agency tiers priced per seat, billed monthly or annually. Cancel any time — your Loops remain exportable as SKILL.md.

Solo
3 Loops, 1 brand profile, personal use. For trying the tool and keeping a personal workflow library.
Free
Start free
Team
Unlimited Loops, 5 brand profiles, version history, team sharing. For 3–10 person marketing / rev-ops teams.
£35/seat/mo
Start 14-day trial
Agency Most popular
Everything in Team, plus unlimited brand profiles, client-separated workspaces, audit log, SSO. For 10–30 person agencies.
£79/seat/mo
Book a call
Founder
First 50 agencies only. Lifetime price. Everything in Agency + design partner access + direct feature input.
£25/seat, forever
Claim 1 of 37 left

Questions, answered plainly.

What actually is a Loop?
A Loop is a tested, versioned, brand-tuned AI workflow — authored once in Loops and deployed to every platform your team uses. Under the hood it's a SKILL.md file built on Anthropic's open Agent Skills standard, so the same Loop runs in Claude Desktop, Claude Code, ChatGPT, and Codex.
Do I need to know how to write prompts?
No. Loops is designed for marketing, ops, and client-services roles — not prompt engineers. You describe what the workflow should do in plain English and give it 2–3 examples of outputs you'd consider "good." Loops handles the rest.
How is this different from Anthropic's built-in skill-creator?
The built-in skill-creator is excellent for individual developers making single skills in Claude Code. Loops is built for teams: team-shared workspaces, version history, eval testing against your reference outputs, per-client brand profiles, an audit log, and one-click deployment to multiple platforms — not just Claude.
What does "tested against our good examples" actually mean?
You paste 3–5 outputs your team considers on-bar. Loops runs the authored skill against the inputs that produced them, compares against your references, and flags drift — tone, structure, terminology, banned phrases. Tuning is automated; you approve the final version.
Who owns the Loops my team authors?
You do. Loops are portable. Every Loop is exportable as a plain SKILL.md file — if you ever stop using Loops, your authored workflows go with you and continue to run on any platform that supports Agent Skills.
What's the Founder tier, exactly?
For the first 50 agencies who commit before product launch: £25/seat, lifetime. You never pay the £79 Agency rate. In return you get design-partner access, direct feature input, and your logo (optional) on the initial site. 37 of 50 seats are left.
When does it ship?
Private beta rolls to design partners in May 2026. Public beta July 2026. The waitlist is the fastest route in — Founder-tier holders get first access.

Every teammate, running your best work.

Reserve a founder seat or join the public waitlist. We'll email when your tier opens. No spam, no marketing drips — just product updates and your invite.

Founder seats claimed 13 / 50
Please enter your name.
We only email about your invite and product milestones.
Please enter a valid work email.
Please select a team size.

You're on the list.

We'll reach out within 48 hours with next steps. If you selected Founder tier, expect a short note to schedule a 20-minute call.