kept.DemoTry it free

You did useful AI work once. Here's how kept. makes it reusable.

Follow one rough prompt — branded docs for a new SaaS app — through eight steps: from a messy paste to a reviewed, versioned, reusable bundle. Everything here is a worked example.

1

Project Dashboard

Every project keeps a timeline of the AI work that happened on it — so nothing useful gets lost in a chat window.

Project: kept.

  1. Today 9:12 AMImported ChatGPT conversation
  2. Today 9:14 AMSuggested reusable bundle
  3. Today 9:16 AMCreated Documentation Bundle v1
  4. Today 9:20 AMGenerated 5 outputs
  5. Today 9:24 AMReviewed output for unsupported claims
  6. Today 9:27 AMCreated Documentation Bundle v2
2

Imported AI Work

You paste the rough prompt you already use. kept. reads it and works out the repeatable workflow underneath.

Please create branded documentation for each of the following areas:
- onboarding checklist - what needs to be done on my end for new customers?
- ops troubleshooting checklist - quick fixes in order of likelihood
- network architecture diagram - include ports, logic, data flow
- security overview - for when customers inevitably ask
- pricing - how are we going to monetize this?

Detected repeated workflow: New App Documentation

Possible outputs

  • Customer Onboarding Checklist
  • Ops Troubleshooting Checklist
  • Architecture / Data Flow Diagram
  • Security Overview
  • Pricing Strategy

Missing inputs

  • Product name
  • Target customer
  • Hosting environment
  • Authentication model
  • Data stored
  • Integrations
  • Pricing assumptions
3

Convert to Bundle

Fill the missing inputs once. kept. captures them as reusable project context, so every output is grounded in the same facts.

Project context

Product namekept.
Short descriptionReusable AI workflow system
Target usersPower users, developers, consultants
Hosting / infrastructureAzure Container Apps
Authentication modelEmail/password, NextAuth.js
Data stored / processedPrompts, runs, context blocks
IntegrationsAnthropic, OpenAI, Google Gemini
Support responsibilitiesEmail — hello@usekept.ai
Pricing assumptionsFree tier (20 runs), BYOK for power users
Brand toneDirect, professional, no filler

Outputs configured

  1. 1Customer Onboarding Checklist
  2. 2Ops Troubleshooting Checklist
  3. 3Network Architecture Diagram Spec
  4. 4Security Overview
  5. 5Pricing / Monetisation Draft
4

Generated Outputs

One run produces the whole bundle — five coordinated documents from the same project context.

Run:#1

Bundle:New App Documentation Bundle

Version:v1

Model:Claude Sonnet

Started:9:18 AM

Completed:9:20 AM

Status: Complete

Customer Onboarding Checklist

Before go-live: provision the workspace, confirm the customer's email domain, and share the BYOK key setup guide. Day one: the customer signs in, adds an API key (or uses free runs), and saves their first rig…

View full output

Ops Troubleshooting Checklist

1. Runs failing immediately → check the provider API key in Settings; an invalid key returns a 401. 2. "Free runs used up" → the managed key budget is exhausted; add a BYOK key. 3. Slow responses → check provider status…

View full output

Network Architecture Diagram Spec

Browser → Next.js app (Azure Container Apps, :443) → Postgres (:5432, private). Outbound to provider APIs (Anthropic/OpenAI/Gemini) over TLS. Auth via NextAuth session cookie. Ports & protocols table included…

View full output

Security Overview

Hosted on Azure Container Apps. API keys are encrypted at rest (AES-256-GCM). Sessions are JWT-based. Data stored: prompts, runs, context blocks. SOC 2 certified. Customer data is never stored…

View full output

Pricing / Monetisation Draft

Recommended: usage-anchored freemium. Free tier of 20 managed runs to reach first value, then BYOK for power users (no token markup). Paid tiers gate team features and analytics. Three models compared…

View full output
5

Review Output / Fact Check

Before anything ships, kept. checks the output against the context you provided — and flags claims it can't support.

Review Output · Security Overview

Issue 1 — Unsupported claim

Output says "SOC 2 certified." Context only confirms Azure hosting.

Issue 2 — Missing caveat

Output says "customer data is never stored." No retention policy was provided.

Issue 3 — Inconsistent wording

Authentication described as both "magic link" and "OAuth-only."

Suggested rig improvement

Add instruction: Do not claim certifications, security controls, or data-retention policies unless explicitly provided in context.

6

Version History

Every improvement is a new version, with the run feedback that motivated it — so you always know what changed and why.

Documentation Bundle

v1

Original converted prompt

9:16 AM

Run feedback: Needed edits

v2

Added missing-context handling and unsupported-claim guardrails

9:27 AM

Run feedback: Useful

7

Handoff

Carry the project's decisions into any other AI tool. kept. packages the context so you never re-explain yourself.

Project: kept.
Goal: Improve the New App Documentation Bundle.
Current state: Bundle v2, post-output review.

Important decisions:
- Product name: "kept." (lowercase, with period)
- Positioning: reusable-work system for AI
- Avoid unsupported compliance claims
- Outputs should be customer-facing but honest

Next prompt:
Please improve Documentation Bundle v2 using the project context above...
8

Rerun for New Product

The bundle is now a reusable asset. Point it at a different product, and you get the same five documents — grounded in the new context.

Project context

Product nameCaterease API
Target usersCaterease customers and integration partners
HostingAzure App Service
AuthenticationJWT bearer tokens

You did useful AI work once. kept. made it reusable.

Start free