Local-first prompt intelligence

See the structure. Shape the prompt. Keep the history private by default.

Measurably turns six prompt-structure signals into local coaching across the web, supported Chromium sites, and local Codex and Claude Code sessions today—with a desktop companion moving through platform validation.

Coach a prompt now See what ships today Download Everywhere previews
Six-dimension scoringOne-click local enhancementOn-device voice where supportedDerived-only analytics

Release truth, not roadmap theater

What you can use—and what we are still proving.

Cross-platform source is not the same as a trusted publisher-signed, installed, physically accepted product. We show the difference.

Available now

Web prompt coach

Score, reshape, copy, and continue a prompt locally without installing anything.

Developer preview

Browser workflow

A checksummed Chromium package now offers local scoring and prompt actions; account pairing remains paused.

CLI available

Terminal agents

A checksummed local MCP CLI connects Codex and Claude Code; macOS is locally accepted and cmux inherits host settings.

Engineering preview

Desktop app

Source now includes an encrypted private prompt library, keyword recall, and optional local vector recall; trusted signed downloads are still gated.

Browser 0.8.0 and CLI 0.3.0 previews are downloadable now. Bring explicit next-prompt previews to supported Chromium sites or local MCP coaching to Codex and Claude Code; verify each published SHA-256 receipt before installing.

Install the preview

Try the local prompt coach

Rate it. Rewrite it. Keep moving.

Score all six AES dimensions, enhance or reshape the prompt, dictate on supported browsers, and prepare a current-prompt follow-up—all without sending prompt text to us.

AES evaluates prompt structure with deterministic heuristics. It does not measure model-output quality or prove productivity gains.

prompt_scorer.live
Try:
0AES

Type a prompt to see your
AI Effectiveness Score

Scoring runs locally in your browser.
Nothing leaves this page.

A measurement layer, not another chatbot

Make the request inspectable. Judge the response yourself.

Models, IDEs, and interfaces keep changing. Measurably surfaces context, constraint, and output-format signals you can review before sending, while the resulting response remains the real test.

Read the scoring methodology

A score you can act on

One AES score, backed by six dimensions that explain what to improve: specificity, context, structure, iteration, efficiency, and output definition.

Coaching at the point of work

See the missing context before you send, reshape a prompt in one click, dictate on device where supported, and prepare the next prompt with one shortcut.

Progress without prompt surveillance

Build useful trends from derived scores, tool names, and timestamps. Raw prompt text is not an analytics event.

A memory you choose

The desktop source can save named prompts in a separate encrypted vault and recall them by words or local embeddings—only after an explicit action.

Useful measurement without prompt surveillance.

The privacy boundary is architectural: local scoring is the default path, and prompt bodies are excluded from persisted measurement analytics, logs, and dashboards.

Local by default

Deterministic scoring can run without sending the prompt to a model provider.

Derived history

Store the score, dimensions, source tool, and time—not the sentence that produced them.

Explicit private memory

Saved prompts and embeddings use a separate encrypted desktop vault and never become measurement analytics.

AI only when invoked

A rewrite or deeper model analysis is a separate, explicit action with its own processing boundary.

Inspectable contracts

Versioned schemas and privacy tests make the intended data boundary reviewable.

Community release under evaluation

Trust belongs in the source, the tests, and the release evidence.

Measurably is private and has no approved root license today. Any future public release remains gated on an approved license, governance, reproducible packages, and verified release evidence.

Request a launch notice
Now

Local web and browser scoring, prompt actions, next-prompt previews, on-device voice where supported, and a downloadable MCP CLI for Codex and Claude Code.

In validation

Encrypted explicit-save prompt library, local semantic recall, authenticated local agent, GUI companion mode, browser native pairing, and broader desktop integrations.

Release gate

Trusted publisher-signed installers, macOS notarization, and physical acceptance on macOS, Windows, Linux, x64, and ARM64.

Enterprise and public-sector design partners

Measure prompt structure with explicit data boundaries.

We are evaluating possible controlled pilots for organizations that need local processing and reviewable data boundaries. No certified or production organization deployment is offered today.

Possible pilot scope

Enterprise

Evaluate a measurement policy, retention boundary, rollout plan, and support model. This is not a current production deployment offering.

  • Organization-level derived metrics and controls
  • Deployment and identity architecture review
  • Commercial support and pilot terms

Possible pilot scope

Government

Evaluate mission constraints, local operation, auditability, and evidence needs. This is not a certified or authorized government deployment.

  • Local-first and disconnected-use architecture review
  • Data-flow, threat-model, and release evidence package
  • Procurement and support terms scoped per engagement
Discuss a design-partner pilot

Plans

Start with your own prompts. Scale with evidence.

No credit card is required for the free plan.

Free
$0forever

Measure your own prompting practice.

Start free
Local six-dimension AES scoring
Live web scorer
Personal trend history
Core improvement guidance
Individual plan
Pro
$9/month

Deeper coaching for an individual practice.

Everything in Free
AI-assisted scoring and rewriting
Full personal history
Weekly insight digest
Expanded coaching library
Effectiveness reports and exports
Goals, streaks, and personal records
Organizations
Custompilot

Possible enterprise and public-sector pilot scope; no current certified deployment.

Discuss a pilot
Controlled pilot scope
Data-flow and deployment review
Organization measurement policy
Commercial support terms
Evidence package scoped per engagement

Questions worth asking.

What is the AI Effectiveness Score?

AES is a 0–100 composite of six structural-signal dimensions: specificity, context, structure, iteration, efficiency, and output definition.

Does Measurably collect my prompts?

Raw prompt text is not treated as analytics. Local scoring can run on-device. If you explicitly request an AI rewrite or deeper model analysis, the prompt is processed for that request under a separate boundary.

Can I use the browser extension today?

Yes, as a checksummed developer preview loaded unpacked in Chrome or another Chromium browser. Local scoring and prompt actions work without an account; secure pairing, cloud sync, and public browser-store publication remain paused.

Can I use Codex, Claude Code, or cmux today?

Yes on the locally accepted macOS path. Install the checksummed CLI, explicitly configure its local MCP server for Codex or Claude Code, restart the host, and invoke a tool to prove execution. Windows and Linux are packaged source-preview targets pending native host acceptance. cmux inherits host configuration; Measurably does not automatically read chat history, inspect panes, or submit prompts.

Can I download the desktop app today?

Not yet. The macOS, Windows, and Linux architecture is in active engineering validation. We will not call it released until trusted publisher-signed packages, macOS notarization, and physical platform acceptance are complete.

Can Measurably remember useful prompts?

The desktop source candidate now has an explicit encrypted prompt library with keyword recall and optional semantic recall through a user-configured local Ollama or OpenAI-compatible embedding model. It never indexes prompts automatically, and it is not yet a trusted signed public download.

Is the project open source?

Not today. The repository is private and unlicensed. Any future community release remains gated on an approved license, governance, reproducible packages, and verified release artifacts.

Do you offer enterprise or government licensing?

We are evaluating possible design-partner pilots. No production organization deployment, government authorization, or compliance certification is offered today; possible scope and terms would be agreed per engagement.

Make better prompting a measurable practice.

Try the live score now, or talk with us about shaping the cross-platform community and organization releases.

Try Measurably Download the previews Talk to the team