Describe the job.
Write one sentence: code review, growth teardown, landing-page rewrite, or a new system prompt.
Twenty specialised agents analyse, engineer, assemble, and validate the pack for you — entirely on-device, in under two minutes.


Output
System prompt, developer rules, examples, schema, and scoring — ready to paste.
Keep the idea brief. MetaPromptLab handles role design, context packing, structure, examples, and evaluation rules behind the scenes.
Pipeline summary
The app assembles a reusable pack, not a one-off paragraph. That means stronger output, easier iteration, and cleaner handoff between models.
Write one sentence: code review, growth teardown, landing-page rewrite, or a new system prompt.
Twenty specialists break the prompt into analysis, engineering, structure, examples, and evaluation rules.
Copy a production-ready Prompt Pack with system prompt, developer rules, examples, output schema, and scorecard.
Every agent owns a small job. The handoff stays deterministic, local, and transparent enough to inspect.
Runtime
$ mpl run "code reviewer for python" 20 agents complete · quality score 92 · 13 validation gates passed system + developer + examples + schema + evaluation criteria
Intent, entities, tone, and structure are pulled apart before any wording gets locked in.
The app defines role, context, few-shot angle, and model-facing instructions.
Constraints, format, token budget, and output schema become concrete deliverables.
Agents
20
Phases
5
Runtime
~100s
Cloud
0 calls
Every draft is tightened for redundancy, edge cases, and consistency.
The final pack ships with validation rules you can reuse anywhere.
Each screen is built to move you from raw request to reusable prompt, without detouring through settings, chat logs, or account scaffolding.

The Create tab pulls apart intent before it writes a single instruction.

Analyze, Engineer, Build, Quality, Deliver — sequenced locally on iPhone 15 Pro and newer.

Prompt Packs stay plain text, so they drop into ChatGPT, Claude, Gemini, Copilot, Grok, or Midjourney without cleanup.

The privacy promise is architectural: Apple Intelligence handles the work on-device, with no server layer in the middle.

Free covers 5 runs daily with 8 agents. Pro unlocks the full 20-agent pipeline, exports, and Vault access.

The same prompt-engineering flow expands into a calmer, desk-friendly workspace with more room for Vault, history, and multi-column review.
The privacy promise is not a policy layer added after the fact. It is how the product is wired: local runtime, no accounts, no telemetry, no cloud relay.
Accounts
None
API keys
None
Telemetry
Zero
Cloud
Never
Start free with 8 of 20 specialists and 5 runs daily. Upgrade when you want the full pipeline, exports, and Vault.
5 runs daily · 8 specialists
No card required
Unlimited runs · billed monthly
Cancel anytime
About £2.41 per month
Best recurring value
Pay once · keep it forever
No renewals
SEO Blog Generator add-on
One-time purchase
Everything most people want to check before they install, upgrade, or move their prompt workflow onto the phone.
01 · COMPATIBILITY
Does MetaPromptLab work offline?
Yes. The full pipeline runs through Apple Intelligence on-device after installation.
02 · DEVICES
Which iPhones are supported?
iPhone 15 Pro and newer, running iOS 26 with Apple Intelligence enabled.
03 · INTEROP
Which AI tools can I paste Prompt Packs into?
Any tool that accepts structured prompts: ChatGPT, Claude, Gemini, Copilot, Midjourney, Grok, and more.
04 · PRIVACY
Do you collect prompts or telemetry?
No. There is no account system, analytics pipeline, or cloud prompt relay.
05 · OUTPUT
What is inside a Prompt Pack?
A system prompt, developer instructions, examples, output schema, validation criteria, and a token-budget recommendation.
06 · PRICING
What changes when I upgrade?
Pro unlocks all 20 specialists, unlimited runs, JSON or Markdown export, Vault, and priority processing.
Still curious? Contact support
MetaPromptLab turns a vague sentence into a portable prompt system you can keep reusing across tools.
Trust markers