One-click migration of chat history and memory across ChatGPT, Claude, and Gemini to avoid platform lock-in.
Problem
Users increasingly rely on chat AI as life and work assistants, but chat history and memory live only within one product. Switching to another subscription means losing all past context, making it costly to retrain the AI to understand them. Existing open-source solutions are either closed-source and paid, or outdated and require heavy fixes.
Solution
CarryOn is an open-source MCP server that aggregates chat history from ChatGPT, Claude, and Gemini, and exposes keyword search and context injection so any AI service can access cross-platform memory. Users deploy it to their own device (HAOS or Docker), then add the custom MCP server to their current AI service for seamless use.
Helps indie developers and small teams diagnose why activation emails land in spam and get actionable fixes.
Problem
Indie developers sending product activation emails find that anything with a link gets bounced to spam, and trying plain text, HTML, or no hyperlink doesn't help. They worry about domain health but don't know where to start. Existing tools are either built for large ESPs or give vague spam scores without concrete fix steps for small teams.
Solution
InboxShield is a self-serve email deliverability diagnosis tool: users enter their sending domain, DNS records, and email sample, and the system automatically checks SPF, DKIM, DMARC, blacklists, link reputation, and content triggers, then generates a prioritized fix list. Weekly re-tests push domain health score changes to help teams maintain sender reputation.
Generation and compliance checking for people who write official documents
Problem
People working in government and state-owned enterprises write and review large volumes of official documents with strict national formatting and wording rules, making manual checks time-consuming and error-prone. A non-programmer with twenty years in state-owned enterprises built a national-standard-based document generation and inspection tool with Claude, showing the need is real and urgent.
Solution
Offer document template generation, formatting and wording compliance checks, national-standard rule validation, and revision suggestions. Let users import existing documents, get an issue list, and export a corrected version.
Cut the tokens coding agents spend reading tool output
Problem
Coding agents burn large amounts of tokens reading command and tool output, raising costs and crowding the context window. One Show HN specifically built a tool to cut tokens spent on tool output, showing this waste is common in agent workflows. Another project counts excess code units added by AI models, reflecting broader sensitivity to AI workflow costs.
Solution
Add a local filter between the agent and the terminal that summarizes and deduplicates long logs, test output, and directory listings, sending only key lines to the model. Support per-tool rules and show tokens and dollars saved.
A minimal cross-platform SSH client focused only on vault and password management
Problem
Existing SSH clients like Termius and Tabby are bulky and memory-hungry, yet daily use often only needs vault and password management. Termius charges just to sync the vault, while free tools lack the same UX, which led one developer to build a 76 MB native client. The frustration is real and specific.
Solution
Build a lightweight client focused only on vault, password management, and multi-project sessions, keeping memory under a hundred MB with local encrypted storage and optional self-hosted sync. Skip terminal emulation perfection and nail the vault experience.
Measure and reduce the excess code AI agents pile into your repo
Problem
Existing benchmarks measure solve rate and token cost but do not punish excess code, which makes a codebase harder to maintain long term. One Show HN explicitly created an open-source benchmark to count how many coding units AI models add, showing this is now taken seriously.
Solution
Connect to repo history, quantify the syntax units, duplicate patterns, and mergeable logic added by each AI commit, then surface slimming suggestions and a lean score. Developers can see at a glance which agent is bloating the codebase.
A structured review view for codebases heavily modified by AI agents
Problem
AI agents can now make sweeping changes across a repo, leaving developers feeling out of control about what changed and whether to accept it. One Ask HN asks how to review code in the new AI world, while another notes that PR review was already hard and now that AI writes and reviews code, they get even less feedback than before.
Solution
Aggregate agent-generated diffs into semantic units, generate collapsible change summaries, flag risky edits, and let users accept or reject block by block. The core is helping humans grasp what an agent did in minutes instead of reading raw diffs.
Make nuclear, diamond, and 6C battery progress legible to normal people
Problem
Information on betavoltaic, nuclear diamond batteries, 6C cells, and similar breakthroughs is scattered and highly technical. Ordinary readers and hobbyist investors can't tell real progress from marketing. A community member even asked directly what stage this is all at, with no good aggregation site to point to.
Solution
A vertical tracker that catalogs companies, products, papers, and timelines in nuclear and next-gen energy storage, explaining principles and maturity in plain language, with RSS/email updates and comparison tables.
Find accidentally pasted secrets in Claude Code, Codex, and Cursor sessions
Problem
Developers routinely paste API keys, DB passwords, and private tokens into AI coding sessions. These linger in local session history and leak the moment sessions are shared, logs exported, or devices change. There's no unified scanner across tools.
Solution
A local CLI / menu bar tool that reads local session stores of common agents like Claude Code, Codex, and Cursor, detects suspected secrets using configurable pattern matching plus entropy checks, and reports the session, context, and suggested rotation or redaction.
Catch unsafe types, duplicate code, and junk tests in AI-written TS
Problem
Many projects now generate TypeScript with AI. The code often runs but is full of unsafe type assertions, duplicated logic, and low-value tests that are hard to systematically catch in review. Manually scanning for these patterns is tedious and error-prone.
Solution
A rule set built on oxlint (or an ESLint plugin) specifically targeting AI-generated code patterns: detects unsafe types, duplicate code blocks, and hollow tests, and suggests fixes. Runs locally or as a CI step with zero workflow overhead.
Turn endless agent terminal output into a navigable Markdown workspace
Problem
When a dozen coding agents run in parallel, each spews walls of suggestions, comments, and reasoning into an 88-column terminal. It all scrolls away the moment you run the next command, so nothing can be reviewed, compared, or reused. Developers are left scrolling and losing context.
Solution
A local-first Markdown workbench: agents write .md files into a watched folder via a simple convention or CLI; the app incrementally renders them, indexes content, and lets you search, group by agent/session, cite anchors, and diff revisions — all offline.
Make editing AI-generated 3D print models painless
Problem
A user designs 3D print models with GPT plus Blender MCP and OpenSCAD CLI; results are good but not great, and iterating is painful because every change requires re-describing, regenerating, and re-checking with no version or parameter tracking.
Solution
An AI modeling iteration helper for 3D printing hobbyists that breaks models into parametric parts and reusable edit commands, records parameters, previews, and exports on every change, and supports one-click rollback and A/B comparison.
Translate rental, insurance, and cancellation clauses into plain risk cards
Problem
A user in Europe paid in full with full insurance but was only 1 hour and 1 minute late to pick up the car; the booking was cancelled, the car rented to someone else, and no refund was given. Such clauses are scattered across sites, multi-lingual, and worded to hide the boundaries from ordinary people.
Solution
A rental car risk decoder: users enter the company, country, and pickup/return times, and it fetches and structures key clauses (late grace period, cancellation policy, insurance coverage, fuel policy), then shows plain-language risk levels and tips.
Let Claude Code and Codex share one project memory
Problem
The same users subscribe to Claude, Codex, Grok, Devin, and more, but each tool works in isolation and project context is not shared; a Hacker News Show HN project explicitly offers 'Claude Code and Codex in one workspace, with shared memory', confirming cross-tool memory is real friction.
Solution
A local workspace that stores project notes, decision records, common commands, and task summaries in a standard format, so different agents load the same memory on start and write new conclusions back after tasks.
Let AI use your logged-in browser to read common sites cheaply
Problem
Existing AI browser control tools handle logged-in sites like Bilibili, Zhihu, and Xiaohongshu poorly: reading pages back and forth, failing to find content, being very slow, and blowing up tokens with messy design; one developer ended up hand-writing extraction logic for 30+ sites.
Solution
A browser extension plus local CLI that wraps common sites into structured commands, extracts data with in-page JS to avoid dumping full HTML into context, and supports functional composition and piping to turn multi-step browsing into reusable scripts.
Turn your coding agent's diffs into human-readable change notes
Problem
V2EX users report shipping hundreds of thousands of lines and tens of thousands of commits with AI, across many projects, yet it's hard to quickly understand what an agent changed and why; a Hacker News Show HN project explicitly targets 'plain-English notes on what your coding agent changed', confirming diff review is no longer enough.
Solution
Connect to a Git repo or agent workspace and generate a structured digest after each commit or task: grouped by feature and file, flagged for risk, highlighting changed public interfaces and likely regressions, with daily or per-project changelog rollups.
Upgrade Codex/Claude-generated prototypes into professional-looking designs
Problem
When developers use Codex for UX prototyping, the results often look amateurish—bad fonts, spacing, and colors, with stiff interactions. Even attaching skills like apple-design doesn’t save it. People want a polished look but end up retrying blindly.
Solution
A prototype post-processor: takes AI-generated HTML/CSS prototypes and automatically re-skins the visual layer using curated design systems (font pairings, spacing rhythm, color palettes, micro-interactions), producing a noticeably more professional result. Users can switch styles (minimal, Apple-like, Material) and export design tokens for development.
Manage parallel tasks across multiple coding agents like email
Problem
When developers run tasks across Codex and Claude in parallel, windows scatter everywhere. It’s hard to see which tasks finished, which are stuck, and which need human input. Some are already dragging 5 to 8 windows around, killing productivity.
Solution
A unified agent task inbox: aggregate sessions and tasks from different coding agents into one board. Each task shows status (running / awaiting confirmation / done / failed), with one-click jump to the corresponding agent window. Similar tasks can be merged or serialized.
A local pre-execution check that approves, flags, or blocks coding-agent commands
Problem
Developers letting coding agents run terminal commands are flying blind—a destructive rm -rf or privilege escalation can wipe out hours of work. Existing safeguards either rely on cloud review or require heavy sandbox setup, leaving a gap for a lightweight, purely local gatekeeper at the command level.
Solution
A local interception layer: before any agent tool call executes, it uses a small local model plus rules to judge risk and output approve, send-for-review, or block. Users can customize allowlists and sensitive-path rules, and all judgment runs offline—no code or commands leave the machine.
Search your local videos with one sentence and jump straight to the hit
Problem
Creators and everyday users hoard massive video libraries on disk. Finding a specific moment—‘sunset by the sea’ or ‘waiting for a train’—means relying on filenames or scrubbing timelines, which is painfully slow. Open-source tools exist for local media search, but most people don’t want to tinker with installs and indexing setups.
Solution
A minimal desktop app: point it at a video folder and it indexes in place, read-only. Search in natural language and jump directly to the matching second. Optimized for common footage like gameplay highlights, meeting recordings, and home videos. Global hotkey summons the search bar, and results can be copied as timestamped links.