Track quota reset times and burn rates across Codex, Claude and other AI plans so you stop wasting credits
Problem
Heavy users report that a Codex 5x plan loses 10% of quota after a few questions, and reset times are opaque. One user woke up to find the reset had already happened at 5am, wasting the previous day's quota. Others got charged for a Claude Max upgrade that never activated. People currently rely on forum threads to share reset times, which is scattered and stale.
Solution
A lightweight quota dashboard where users log each AI service's plan, remaining quota and reset cycle manually or via a browser extension. The system predicts the next reset and sends reminders, while aggregating community-reported reset events into a crowdsourced calendar. It also suggests how to burn remaining quota before reset.
Scan MCP tools for prompt injection, leaked keys and overreach before you install them
Problem
The MCP ecosystem is exploding, and more servers are published by solo devs or unknown third parties. Tool descriptions can hide prompt injection, hardcoded secrets or over-privileged access. When developers plug a third-party MCP into Claude Desktop or their own agent, there is almost nothing that warns them the tool description could trick the model into leaking data. Existing projects only do directory scanning with limited coverage.
Solution
A CLI and GitHub Action that parses an MCP server's tools descriptions and input schemas, then uses a rule library plus a local small model to detect prompt injection patterns, sensitive field exfiltration, suspicious URLs and hardcoded tokens. It outputs a risk report with fix suggestions and supports CI diff alerts on MCP changes.
Turn idle kids' audio hardware into a client for your own audiobook server
Problem
Parents buy Yoto-style players, kids outgrow them, devices get shelved — while the parents already run Audiobookshelf. The two don't talk, so someone had to write their own bridge with knob browsing and a 16x16 cover screen. Classic 'home self-hosted media + idle hardware' gap.
Solution
A drop-in home server add-on: connects your self-hosted audiobook library to old players, with knob browsing, cover art, favorites, and offline sync. Ships with flashing guides and step-by-step tutorials to keep non-geek parents from bouncing.
Add do-not-disturb rules and auto-approve allowlists to cross-session AI coding tools
Problem
Multi-session AI coding tools fire 'held message from another session' approval prompts that break workflow. Users complain it's 'killing my output' but can't fully disable it — they either deny or approve one by one. Chat assistants also keep asking 'want to know more?', forcing manual behavior tuning.
Solution
A middle layer or plugin: define allowlists and quiet rules per session/agent, auto-deny low-priority cross-session messages, and batch the ones that truly need a human into an hourly/daily review panel. Includes a one-click AGENTS.md snippet generator.
Turn a failed agent run into a reusable regression test
Problem
Agent failures are flaky and hard to reproduce: one step produces a bad output but you can't tell which step or why. Developers rerun from scratch instead of replaying a single step and testing a fix. Agentfork and similar show 'replay from any step and test fixes' is a real need — but there's no lightweight version for regular developers.
Solution
Record every step's inputs, outputs, and tool calls into a replayable trace. Pick the failing step, tweak the prompt/args, rerun only that step and downstream, diff old vs new trace, and export the case as a regression test. Works with OpenAI/Anthropic-compatible APIs.
Move developers from line-by-line AI review to risk-tiered spot checks
Problem
Trust in LLM-written production code varies wildly: some review every line, some only read tests, some barely look. Threads repeatedly show "I review everything so I can't even use that many tokens" and "I'm not sure how much to trust it." Without a risk-tiered checklist, teams either over-review or miss critical defects.
Solution
A lightweight review companion: pulls the PR diff and generates a tiered checklist by code type (auth, payments, migrations, UI), flagging high-risk functions, missing tests, and suspicious error handling — then says "you only need to look at these three spots." Logs blind spots over time so teams build their own review policy.
For people drowning in AI-generated messages, auto-summarize and mute the low-signal ones
Problem
Bosses now dump AI-generated messages into Slack all day — long, repetitive, action-free. Teammates have developed a reflex to skip them: "whenever I see his automated messages, my natural reaction is: why do I even bother reading that?" As AI-written code and docs become normal, downstream reviewers face more noise, not less.
Solution
A Slack/Teams extension (or local browser plugin) that scores incoming messages for "actionable information density": flags templated AI tone, duplicates, and long threads with no next steps, then collapses them into a one-line summary with the original behind a toggle. Per-sender filter strength and a 'collapse rate' report so teams can see who is generating the noise.
Self-serve compliance checks for ads and crypto/finance content on small sites
Problem
Small sites in crypto, finance, or religious-compliant niches often cannot tell whether their content crosses a line and rely on guesswork. A site already screens 800+ coins against AAOIFI Shariah ratios on HN, showing real demand for vertical compliance screening, yet most small teams cannot afford advisors.
Solution
Offer a self-check tool where users pick business type and region, then get a list of risk items, required disclosures, and a saved record. Focus on explainable rules for niches like Halal finance, advertising claims, and crypto assets, clearly not legal advice.
Use local AI to turn completed tasks into skill tracking and growth reports
Problem
People tracking personal growth waste time figuring out which skills a finished task actually exercised, so they stop logging. A locally hosted AI that maps task descriptions to skills already appeared on HN, confirming the pain, but existing tools are single-user and fiddly.
Solution
Build a lightweight workspace where users log completed tasks and a local AI maps them to skills and goals, generating periodic growth reports and next-step suggestions. Import existing todos/notes and view time invested per skill.
Add AI-crawler paywalls and stats to any small site in minutes
Problem
AI crawlers scrape content at scale without paying, and independent publishers lack both leverage and tooling to identify or throttle them. A pay-per-page USDC scheme for AI crawlers already appeared on HN, showing real demand, but existing options are blockchain-heavy and hard for normal site owners.
Solution
Provide lightweight middleware or reverse-proxy configs that detect common AI crawler user agents, apply paywalls or rate limits per page, and report crawl volume. Support simple billing via API keys or invoices instead of crypto wallets, with custom allowlists and free quotas.
Parse bank SMS into expenses locally, never uploading financial data
Problem
In many regions banks and payment apps only notify users by SMS, so keeping a ledger means manually copying amounts and merchants. Someone already built a local-first expense tracker that parses bank SMS, confirming demand for keeping financial data off the cloud, yet most tools are tied to one bank or country.
Solution
Ship a purely local app that reads bank and payment SMS, parses amount, merchant, and balance with template rules, and auto-categorizes entries across multiple banks. Data stays on device, exports to CSV, and lets users edit parsing rules.
Self-hostable, pay-per-page table extraction from PDFs and images
Problem
Small teams constantly need to turn tables inside PDFs, scans, and screenshots into CSV/Excel, but existing tools either export whole pages or upload files to an opaque cloud. Someone already shipped a free table-extraction tool on HN, which shows steady demand while general SaaS options stay expensive and rigid.
Solution
Offer a web app plus an API that accepts a PDF or image, detects table boundaries, and returns clean CSV/JSON with merged-cell handling and batch mode. Local OCR plus rule-based post-processing lets users review and fix column structure before exporting.
Buy AI design asset credits that never expire — no subscription required
Problem
Many designers and programmers use AI image platforms only in bursty, high-intensity sessions, yet mainstream products bundle monthly subscriptions with credits that reset even if unused, while heavy users run out. Some developers have resorted to building their own platform, confirming the mismatch between usage frequency and pricing is a widespread pain.
Solution
A workspace for generating images and design assets sold as one-time credit packs that never expire, showing the exact cost of each generation. It includes common design-asset templates and a batch generation queue with export to common design tools, so low-frequency users don't pay monthly.
Pixel billboards and a self-serve ad console inside a vibe-coding village game
Problem
Indie developers who ship small tools struggle to reach peers cheaply — standard ad platforms are click-auction based and tiny budgets never get traction. A pixel management game that attracts developers and AI-coding hobbyists, on the other hand, has natural attention and dwell time.
Solution
Open billboard slots in the pixel village game where developers upload a product card and link, rent a virtual plot by the week, and see impressions and clicks in a console. Ads blend into the scene in pixel style and can be tagged to developer-community themes.
Test your AI agents against prompt injection in the browser and get a security report
Problem
Once developers connect AI agents to browsers, file systems and internal APIs, they have almost no way to verify whether a prompt injection can make the agent escalate privileges, leak data or take dangerous actions. Enterprise security tooling is heavy and mostly covers text chat, not tool-using agents.
Solution
Load a controllable simulated agent environment in the browser with a built-in suite of prompt injection and privilege-escalation attacks, run them one by one and produce a risk report: which attacks succeeded, what tool calls fired, what data leaked, plus hardening suggestions. Import your own agent config and tool list.
Practice code review and debugging on bugs actually written by AI assistants
Problem
Code from AI assistants often ships subtle edge-case and logic bugs, and developers have gotten used to pasting and merging fast, letting review skills decay. Existing LeetCode-style problems are hand-designed and deterministic, far from the real defect shapes produced by AI.
Solution
A practice site where each kata is runnable code containing a real bug introduced by AI-generated code; users locate it, edit in-browser, and pass the tests, then get a defect-type breakdown and improvement notes. Filter by language and bug type, and track fix accuracy and time.
Find and free the stray processes and ports your AI coding agents leave on your Mac
Problem
After running Claude Code, Cursor Agent or a local model, many processes never exit — they keep eating CPU, memory and ports, spinning fans and draining battery while the user has no idea what is running. Developers end up scrolling Activity Monitor, unable to tell agent processes from system ones and afraid to kill the wrong thing.
Solution
Scan and identify leftover processes spawned by AI coding agents, Node/Python children, local model servers and MCP servers, grouped by project and start time with CPU, memory and port usage. Stop them safely in one click and see the resources reclaimed, with a menu bar indicator and whitelist to protect running tasks.
One-time phone numbers to receive SMS verification codes, like temp email
Problem
Signing up for overseas services often requires phone verification, but users don't want to expose their real number. Someone asked if there's a temp phone number like temp email, showing the need is real and current options aren't convenient enough.
Solution
A web-based temp phone number service: pick a number and receive SMS codes, with validity from 10 minutes to 24 hours. Supports multiple countries, shows messages in real time, and auto-destroys on expiry.
Automatically pick the right skills and MCP servers for your coding agent based on your project stack
Problem
Every coding agent (Claude Code, Codex, Gemini CLI, etc.) has two issues: training data has a cutoff, so it confidently uses renamed or removed APIs; and there are thousands of skills and MCP servers, making it unclear which are worth adding. One developer built a tool specifically to filter them.
Solution
Scan project files (package.json, requirements.txt, etc.) to detect the stack, then recommend matching skills and MCP servers from a curated catalog with version compatibility notes. Generate ready-to-paste config snippets.
Host all directory submission badges in one place—no more redeploys for a single badge
Problem
When submitting a SaaS to directories, you're often asked to add their badge to your homepage. Each time means opening the editor, editing the footer, committing, pushing, and waiting for deployment—repeated dozens of times. One developer wrote a whole post ranting about this flow.
Solution
A lightweight script or hosted page manages multiple directory badges and injects them dynamically into your site. Add the script once, then toggle or swap badges in the BadgeDock dashboard without redeploying.