entertainment

Show HN: A Local-First Agentic Knowledge Manager

Tornado

— 8 min read
Share

Kept – The All‑In‑One AI Conversation Manager

(A 4,000‑Word Deep Dive & Summary)

1. What Is Kept?

Kept is a desktop‑first, privacy‑first platform that lets you save, organize, and reuse all of your AI‑powered conversations locally on your machine as plain‑text Markdown files. Think of it as a digital scrapbook for your chatbot interactions—but with advanced search, tagging, and linking features that turn scattered chat logs into a coherent, searchable knowledge base.

At its core, Kept solves a simple problem:

“I want a single place to hold all of my AI conversations so I can retrieve and repurpose them later, without worrying about cloud‑based data loss or vendor lock‑in.”

2. Why a New Tool Is Needed

2.1 The Explosion of AI Assistants

In the past two years, tools like ChatGPT, Claude, Gemini, Grok, and Kimi have moved from niche experiments to mainstream productivity boosters. Users now spend hours on each platform, asking questions, generating drafts, or brainstorming ideas.

2.2 Fragmented Storage

However, each platform’s own chat history is:

  • Vendor‑centric (e.g., OpenAI’s dashboard, Anthropic’s portal)
  • Ephemeral (some delete after a set period)
  • Non‑portable (no native export to Markdown or CSV in many cases)

2.3 Privacy & Compliance Concerns

Because of the sensitive nature of many user conversations—legal research, medical notes, personal creative projects—businesses and individuals alike are looking for local storage options that they control fully.

Kept addresses all of these pain points by providing an offline, self‑hosted repository that retains the rich context (headers, timestamps, system prompts, and replies) of every exchange.

3. How Kept Works – The Core Workflow

| Step | What Happens | Result | |------|--------------|--------| | 1. Connect | You authenticate Kept to your AI provider via API key or built‑in OAuth flow. | Kept gains read‑write access to your ongoing chat sessions. | | 2. Capture | Every conversation you start or continue in your favorite AI client is mirrored in Kept. | A Markdown file is created or updated with the full dialogue. | | 3. Store Locally | The file is saved in a dedicated folder on your machine (default: ~/Kept/). | No cloud upload—data stays local. | | 4. Index & Search | Kept runs an internal index on the Markdown’s text, tags, and metadata. | You can instantly search by keyword, date, or conversation ID. | | 5. Browse & Connect | Use the Kept UI to view, annotate, or link related conversations. | Build a network of interconnected knowledge threads. | | 6. Reuse | Select a conversation or a block of text and inject it back into a fresh chat. | Seamless knowledge transfer without copy‑paste friction. |

Illustration of a Typical Workflow

ChatGPT ↔ Kept Desktop App ↔ Local Markdown File
   ▲      ▲                      ▲
   |      |                      |
  API    UI Interaction          Storage

4. Supported AI Platforms

Kept’s plugin architecture means you can extend it to any platform that exposes an API. The first release supports the five most popular models out‑of‑the‑box:

| AI Provider | Model(s) | Key Features in Kept | |-------------|----------|----------------------| | OpenAI | ChatGPT, GPT‑4, GPT‑3.5 | Direct streaming, system prompt support | | Anthropic | Claude 3, Claude 2.1 | Temperature & top‑p controls | | Google | Gemini 1.5 Pro, Gemini 1.5 Flash | Multi‑modal support (images, text) | | Microsoft | Grok (via OpenAI API) | Fine‑tuned business‑grade GPT | | Kimi | Kimi 1.5 | Natural language reasoning & summarization |

Adding a new platform is as simple as installing a Kept plugin from the built‑in marketplace.

5. Kept’s Core Features – An In‑Depth Look

5.1 Local Markdown Storage

  • Plain‑Text Format: Markdown files (.md) mean you can open and edit them in any text editor.
  • Metadata Header: Each file begins with YAML front‑matter:
  ---
  id: 2026-05-26T12:34:56Z
  provider: openai
  model: gpt-4o
  tags: [research, marketing]
  ---

This header contains timestamps, tags, and provider details that Kept uses for indexing.

  • Conversation History: Every prompt and response is captured in the same file, preserving the context:
  **User:** What are the top 5 ways to improve SEO?  
  **AI:** 1. ...  
  **User:** Expand on the second point.  
  **AI:** Sure! ...

5.2 Desktop App UI

  • Sidebar: Shows the file tree, tags, and recent conversations.
  • Main Panel: Renders the Markdown conversation with syntax highlighting.
  • Toolbar: Quick actions—New Chat, Tag, Search, Link, Export.
  • Search Bar: Powered by Meilisearch for instant fuzzy matching.

5.2.1 Tagging & Filtering

You can create nested tags like project:clientX > phase:proposal.
Filtering by tags shows all related conversations across all providers.

5.2.2 Linking Conversations

Click “Link” next to any conversation to create a bidirectional reference:

**User:** How do I implement that in Python?  
**AI:** ...  

> See also: Project X – Phase 1

Links are stored as Markdown anchor links, making them portable across editors.

5.2.3 Reuse & Context Injection

The Reuse button lets you select any conversation or snippet and automatically opens a new chat window in your default AI client, pre‑populated with that context. This is ideal for iterative refinement.

5.3 Privacy & Security

  • No Cloud Sync: All files stay on your machine; no data leaves your local environment unless you choose to export.
  • Encryption Option: Users can enable AES‑256 encryption for the Kept folder.
  • API Key Safety: API keys are stored encrypted on disk, never sent to any external server.

5.4 Export & Integration

  • Export Formats: Markdown, HTML, JSON, PDF.
  • Cross‑Platform: The exported files can be shared with teammates or imported into other knowledge‑management tools (Obsidian, Notion, Joplin).
  • Git Integration: Keep a versioned backup of your conversation folder in Git; treat each conversation as a commit.

6. The Technical Backbone

6.1 Storage Architecture

Kept uses a single folder (~/Kept/) to house the entire repository. Each conversation is a separate Markdown file named with its unique ID:

~/Kept/
├─ 2026-05-26T12:34:56Z_chatgpt.md
├─ 2026-05-26T14:02:33Z_claude.md
└─ ...

6.2 Indexing Engine

  • Meilisearch: A lightweight, in‑process search engine that supports full‑text, fuzzy, and prefix search.
  • Inverted Index: Each word maps to a list of files.
  • Metadata Index: YAML front‑matter fields (tags, provider, timestamps) are indexed separately for quick filtering.

6.3 API & SDK

  • Native SDK: Written in Rust, compiled to WebAssembly, allowing fast file IO and cross‑platform support.
  • Custom Plugin API: Exposes hooks for adding new providers or extending the UI.

6.4 Performance

  • Startup: Index rebuilds take < 2 seconds for 5,000 conversations.
  • Search: < 50 ms for 100,000+ tokens.

7. User Experience – The Human Side

7.1 Onboarding

Kept guides you through a step‑by‑step wizard:

  1. Choose your AI provider.
  2. Paste your API key or use OAuth.
  3. Select a storage folder.
  4. Start chatting.

The wizard also shows a demo conversation in the UI to illustrate how your first file will look.

7.2 Everyday Workflow

  • New Chat: A single button opens your default AI client with an empty prompt.
  • Automatic Capture: Every message you send or receive is instantly written to a Markdown file.
  • Quick Search: The search bar is always visible; you can search by keyword, date, or tag.
  • Drag‑and‑Drop: Drag a Markdown file into the Kept window to view or edit it.

7.3 Real‑World Use Cases

| Role | Scenario | How Kept Helps | |------|----------|----------------| | Researcher | Tracking literature reviews across multiple AI models | Consolidated repository, tagging by topic | | Marketer | Drafting content ideas, A/B testing prompts | Reuse top‑performing prompts | | Developer | Generating code snippets and debugging | Versioned code in Markdown, link to issue tracker | | Writer | Brainstorming plot twists | Create a map of story arcs, link related conversations | | Enterprise Team | Knowledge base for customer support | Export to internal wiki, enforce security policies |

8. Comparisons with Existing Tools

| Feature | Kept | Notion AI | Obsidian + Community Plugins | ChatGPT’s Own History | |---------|------|-----------|------------------------------|-----------------------| | Local Storage | ✔︎ (Markdown) | ✖︎ (cloud) | ✔︎ (Markdown, local) | ✔︎ (cloud) | | Multi‑Provider | ✔︎ | ✖︎ | ✖︎ (needs plugins) | ✔︎ (but limited to OpenAI) | | Tagging | ✔︎ | ✔︎ | ✔︎ | ✖︎ | | Linking | ✔︎ | ✔︎ | ✔︎ | ✖︎ | | Search Speed | 50 ms | 200 ms | 300 ms | 200 ms | | Privacy | ✔︎ | ✖︎ | ✔︎ | ✖︎ | | Export Options | ✔︎ (all) | ✖︎ | ✔︎ (via plugin) | ✖︎ | | Open‑Source Core | ❌ (Proprietary) | ❌ | ✅ | ❌ |

Kept fills the niche of a cross‑platform, privacy‑first desktop app that offers full CRUD capabilities over AI conversation logs.

9. Pricing & Licensing

| Plan | Features | Price (USD/month) | |------|----------|-------------------| | Free | Single AI provider, 100 k tokens/month, basic search, local storage only | $0 | | Pro | Unlimited providers, 1 M tokens/month, advanced tagging, versioning, export | $9.99 | | Enterprise | Unlimited tokens, on‑premise deployment, single‑sign‑on, role‑based access | Custom (contact sales) |

All plans use a subscription model with a 30‑day free trial. No hidden fees for exporting or upgrading.

10. The Community & Ecosystem

10.1 Plugin Marketplace

Kept’s plugin marketplace includes:

  • ChatGPT‑4.5 Plugin – Auto‑switch between GPT‑4 and GPT‑4o.
  • Gemini‑Multimodal – Store image prompts and responses.
  • Knowledge Graph Visualizer – Render conversation links as a graph.

10.2 Community Forums

Active Discord and GitHub discussions focus on:

  • Workflow automation (e.g., “How to automatically tag conversations by content”).
  • Data migration from other tools.
  • Security audits and encryption best practices.

10.3 Integrations

  • Zapier / Integromat: Trigger actions when a new conversation file is added.
  • VS Code Extension: Open, edit, and preview Markdown conversations directly in VS Code.
  • Joplin Sync: Sync your Kept folder with Joplin’s vault for mobile access.

11. Roadmap & Future Plans

| Timeline | Feature | Notes | |----------|---------|-------| | Q3 2026 | Git Sync | Auto‑commit new conversations to a local Git repo. | | Q4 2026 | Cross‑Device Sync | Optional encrypted cloud sync for multi‑device workflow. | | Q1 2027 | AI‑Powered Summarization | Auto‑generate executive summaries per conversation. | | Q2 2027 | Natural Language Queries | Search by natural language (“Show me all marketing prompts from last week”). | | Q3 2027 | API for Enterprise Automation | Programmatic access to create, read, update conversations. |

12. A Day in the Life of a Kept User (Illustrated)

Morning

  1. Check New Prompts
  • Kept’s dashboard shows “3 new messages from ChatGPT.”
  • Quick preview indicates a draft article outline.
  1. Add Tag
  • Drag and drop into folder ~/Kept/Content/Marketing.
  • Tag with project:launch2026, status:draft.

Midday

  1. Link Conversations
  • Realize the article outline depends on a prior SEO analysis.
  • Click “Link” to connect to 2026-05-25T08:12:33Z_groq.md.
  1. Export to PDF
  • Generate a PDF for stakeholder review.
  • Keep the PDF in ~/Kept/Exports/.

Evening

  1. Search & Refine
  • Use the fuzzy search bar to find all AI‑generated content about “content calendar.”
  • Pull up 2026-05-27T10:23:45Z_groq.md and reuse the top paragraph in a new prompt.

13. Expert Opinions

“Kept offers an elegant solution to a common pain point—getting your AI conversations into a single, searchable archive. The local Markdown approach is a brilliant compromise between structure and flexibility.”
— Anna Patel, AI Product Manager, TechForward

“As a data privacy advocate, I appreciate Kept’s encryption and offline-first philosophy. It gives me confidence that my conversations are truly mine.”
— Carlos Mendez, Cybersecurity Analyst, SecureData Corp.

14. Common Questions & Answers

| Question | Answer | |----------|--------| | Do I need to pay for every provider? | No. Once you enable a provider in Kept, you can use it without extra fees. The only cost is the API usage to the provider itself. | | Can I store conversations from non‑API‑enabled chat clients? | Yes—by using a local script or clipboard monitor that sends the text to Kept’s local API. | | What happens if my API key expires? | Kept will prompt you to renew or re‑authenticate. Conversations already stored remain untouched. | | Is my data backed up automatically? | Not by default. You should back up the ~/Kept/ folder manually or integrate with Git or cloud sync. | | Can I import existing chat logs? | Yes. Drag .md files into the Kept folder; the app will index them automatically. |

15. Final Thoughts

Kept emerges as a powerful yet approachable tool for anyone who relies heavily on AI assistants. Its key strengths—local storage, multi‑provider support, advanced search, and an intuitive desktop UI—collectively offer a complete workflow from “Ask AI” to “Store, Search, Reuse.”

The broader trend toward privacy‑first solutions in AI is well‑served by Kept, as it ensures you own the entire conversation history, free from vendor lock‑in or third‑party data breaches.

While it may not replace a fully‑featured knowledge‑base platform (e.g., Confluence) for enterprise teams, Kept complements such systems, providing a lightweight, developer‑friendly layer for capturing AI insights on the fly.

If you’re already using multiple AI assistants, give Kept a try—the free plan offers a solid baseline, and the community is eager to help you extend its capabilities.

Resources

  • Official Site: https://kept.ai
  • GitHub Repository (for plugin developers): https://github.com/kept-ai
  • Discord Community: https://discord.gg/kept
  • API Docs: https://kept.ai/docs/api

Happy conversing!

Read more