🥷 AI Ninja Army — Weekly Intel: When AI Tools Stop Pretending They're Magic
Your weekly dose of niche AI tools the mainstream missed. This is a discovery newsletter, not a tutorial — 7 tools, 5 minutes, no hype.
This Week's Finds
This week I kept running into the same problem: teams building AI workflows without any visibility into what actually works. Some of these tools try to solve it by adding monitoring layers. Others build better scaffolding for the agents themselves. And one reminds us that sometimes the best "AI tool" is just better prompt templates. They're not unified by category — they're unified by accepting that AI doesn't run on faith.
Wellows — AI visibility tracking for LLM citation opportunities
What it does: Monitors where your content gets cited across ChatGPT, Gemini, Perplexity, and other LLM answer engines. Tracks which prompts generate responses that mention your brand, then maps outreach opportunities back to you.
Who it's for: Content marketers at agencies managing multiple client domains. SEO teams that realized Google's LLM pipeline is now a distribution channel worth tracking.
What's interesting: For once, a visibility tool doesn't hide its math behind "enterprise pricing." The free tier gives you 10 keywords to test the concept. The real value hit me at the $97/month tier: you get 6,000 responses analyzed monthly and unlimited outreach tracking. That's enough to see if this channel matters before you commit to the $297 Starter plan.
Honest take: The product assumes you already have content worth citing. It's a monitoring tool, not a content creation tool. If your content doesn't show up in LLM outputs naturally, Wellows tells you that — which is useful data — but won't fix it.
Pricing: Free (KIVA), $37/month (Lite), $97/month (Essential), $297/month (Starter), $497/month (Pro). Per-domain billing.
Manus — General-purpose AI agent for delegated research and multi-step work
What it does: Cloud workspace agent that plans and executes multi-step tasks: research, analysis, coding, slide generation, website scraping. You describe what you need, it breaks it into subtasks and runs them sequentially with access to the web and file systems.
Who it's for: Solopreneurs and small teams outsourcing research or document production. Anyone who currently uses ChatGPT for "make me a slide deck about X" but wants it actually built and deployed.
What's interesting: The credit system is transparent. 4,000 credits at $20/month is roughly 20-40 research tasks depending on complexity. More important: all paid plans include 20 concurrent tasks and 20 scheduled tasks. You can queue work and let it run while you sleep. Daily refresh of 300 free credits means even the free tier isn't completely useless.
Honest take: Agent outputs look finished while containing factual errors. Manus doesn't hallucinate less than ChatGPT — it just does it in a formatted document you'll trust more. Always validate the work. Always.
Pricing: Free (300 daily credits), $20/month (4K credits), $40/month (8K credits), $200/month (40K credits). Annual billing saves 17%.
🔗 https://www.therundown.ai/tools/manus
Basalt — No-code AI project orchestration and evaluation platform
What it does: Central hub for managing prompts, datasets, test cases, and evaluation workflows. Import data or generate synthetic test cases. Score AI responses against predefined criteria. Run human evaluation sessions to refine outputs.
Who it's for: AI teams moving from experimentation to production. Anyone tired of managing prompts in Notion and evaluation results in spreadsheets.
What's interesting: Basalt integrates directly into your codebase and catches errors during production, not just at prototype stage. The platform standardizes how teams evaluate and iterate on AI features, which reduces the "let's tweak the prompt manually for three weeks" death spiral.
Honest take: Pricing starts at €3,500/month with a 3-month minimum. That's enterprise-only territory. No startup pricing. No freemium. If you're running a solo AI project, this isn't for you. Even at that cost, you're paying for structure, not for magic.
Pricing: €3,500/month (Basalt Core, 3-month minimum). After that, cancel anytime.
Macaron AI — Personal AI agent with lifetime free credits
What it does: Build custom mini-apps through natural language prompts. Calorie tracker, portfolio simulator, travel planner, personality test — describe what you want, it creates a functional tool you can use and share.
Who it's for: Non-technical users who want AI tools customized to their life, not their job. People who downloaded ChatGPT, realized it's a search replacement, and want something that actually remembers their preferences.
What's interesting: You get 60 almonds (in-app currency) for free, no expiration. That builds 6 custom apps or funds one month of light tinkering. The Classic plan is $13.99/month for another 60 almonds, which is absurdly cheap. Sweet plan runs $18.99/month for 350 almonds. The only difference: pinned projects and voice interaction limits. Refreshingly honest pricing structure.
Honest take: Macaron is personal-use focused, which means it's not designed for commercial applications or team collaboration. The apps you build are functional but simple. Don't expect a custom tool to replace a full software product.
Pricing: Free (60 almonds lifetime), Classic $13.99/month (60/month), Sweet $18.99/month (350/month). Annual billing available.
🔗 https://macaron.im/blog/macaron-free-vs-pro
Portia AI — Open-source agent framework for controllable, authenticated workflows
What it does: Framework for building production-grade AI agents with stateful execution, tool authentication, and human-in-the-loop gates. Built for high-risk automation where you need to audit every decision. Includes Rezonant, a workspace that turns PRDs and call notes into actionable Jira/Linear tickets.
Who it's for: Engineering teams building internal AI tools. Anyone who's had an AI agent blow up production and now requires review gates before execution.
What's interesting: This is open-source, so you can self-host. The cloud features handle authentication and plan tracking, but you're not locked in. Multi-agent planning with explicit breakpoints is the real differentiator — agents propose plans, humans review, then execution runs.
Honest take: Open-source means you own the code but you own the maintenance too. Cloud pricing isn't published, which always means "enterprise sales call." If you need this level of control, you're probably already comfortable with that sales conversation.
Pricing: Open-source (free). Cloud and enterprise pricing available, not published.
🔗 https://www.fixedlabs.ai/tools/portia-ai
PromptCraft — Free prompt engineering that actually works
What it does: Converts rough ideas into structured prompts. You describe what you want, it returns a prompt with context, constraints, and expected output format. Works with any LLM (ChatGPT, Claude, Gemini, Llama, etc.).
Who it's for: Anyone who's spent 20 minutes tweaking the same prompt in ChatGPT and wondered if there was a better way. Teams standardizing prompt templates across departments.
What's interesting: It's free. No freemium trap. No paid tier lurking behind the "advanced features" button. Your prompts stay local in your browser. The tool doesn't log what you send it (they say — check the privacy policy). For once, someone built a useful tool and didn't immediately try to monetize the conversation history.
Honest take: A good prompt still requires you to know what you're asking for. PromptCraft structures it better, but garbage input still produces structured garbage output. It's a productivity tool, not a strategy tool. Also, the homepage shows "free access to core features" with a hint that "advanced features may be added over time" — so I'd expect pricing eventually.
Pricing: Free.
Dopamine.chat — No-code agent builder with native WhatsApp/Telegram support
What it does: Build AI agents through natural language description. Deploy as chat, task runners, or specialists. Native WhatsApp and Telegram connectors. Agents can debate each other to generate contrasting perspectives.
Who it's for: Small teams and solopreneurs deploying customer-facing bots. Anyone who wants Zapier automation but with actual AI decision-making instead of if/then logic.
What's interesting: The free tier includes 30 credits/month and unlimited agents. That's a real free tier, not a 14-day trial. The Pro plan runs 200 credits/month, which is accessible pricing for testing. The "agent debates" feature is niche but genuinely useful for use cases where you need conflicting analysis (business decision prep, legal review, etc.).
Honest take: No-code agents still require you to understand your workflow well enough to describe it clearly. The tool won't rescue a half-baked automation idea. Also, credit consumption isn't transparent upfront — you have to test to know what your automation actually costs.
Pricing: Free (30 credits/month), Starter (more credits, custom model API keys), Pro (200 credits/month, priority support). Exact Starter pricing not published.
🔗 https://www.blog.dopamine.chat/2026-04-11-dopamine-chat-vs-lindy-ai
🥷 Ninja Pick of the Week
Portia AI. Because you don't need more agent frameworks — you need agents that don't require a postmortem after they run. The human-in-the-loop gates and stateful execution are boring infrastructure that nobody marketing blogs will mention. That's exactly why it matters. Bookmark it before it gets acquired and ruined.
