Product Hunt Weekly 2026-09-25: AI Agents Move into the Whole Workflow
Period: 2026-09-18 to 2026-09-25 Sources: Product Hunt API v2, official product pages, Hacker News
TL;DR: This week’s Top 20 launches push AI beyond answers: finding customers, making videos, handling support, operating computers, and continuing software work from a phone. Jev turns model output into structured decisions, Superset Mobile makes coding agents available away from a desk, and Arcjet tackles runtime security. The practical catch is just as important: permissions, cost, recovery, and pricing still need human judgment.
Top 20 products
| # | Product | Upvotes | Tagline | Category |
|---|---|---|---|---|
| 1 | Ami AI | 625 | Lovable for getting customers | Sales, Marketing |
| 2 | Clueso MCP | 539 | Create and edit videos by chatting | AI, Animation |
| 3 | Jev | 534 | Fast, structured AI decisions | API, Developer Tools |
| 4 | Superset Mobile | 504 | Your coding agents, now in your pocket | iOS, AI |
| 5 | AINA | 497 | Find your job search blind spots with an AI coach | Productivity, Career |
| 6 | ProductBridge | 474 | AI-native customer support and feedback agent | Customer Success |
| 7 | Solid | 458 | Agents with their own computers, accounts, and budgets | Productivity, AI |
| 8 | Sai | 421 | The autonomous computer fleet at your command | Productivity, AI |
| 9 | Naise AI | 409 | Autonomous marketing agents that actually execute | Social Media, Marketing |
| 10 | MosMos | 407 | Voice writing before, during, and after meetings | Writing, Audio |
| 11 | Arcjet | 393 | Secure the AI agents you're building at runtime | Developer Tools, Security |
| 12 | Sider Omni Sidebar | 374 | Give every Mac app an Agent Sidebar | Mac, Productivity |
| 13 | Claude Opus 5.5 | 345 | Anthropic's first model in their new Claude 5.5 family | Developer Tools, AI |
| 14 | Mycel | 338 | Bring one past deliverable. Mycel drafts every future one. | Productivity, SaaS |
| 15 | WZRD | 330 | AI-native documents, slides, forms and sheets that talk back | Productivity, No-Code |
| 16 | Scholé Learn by Building | 321 | Learning should be more than “click next” | Productivity, Education |
| 17 | Ari by Ariso | 315 | AI bar raiser for the most ambitious teams | Task Management, Meetings |
| 18 | Floot MCP | 311 | Build and ship web and mobile apps inside Claude or ChatGPT | Developer Tools, No-Code |
| 19 | Anomalo | 304 | Your data is always talking. Don't miss what it's saying. | AI, Data & Analytics |
| 20 | SereneDB | 261 | Ultra-Fast Search & Analytics Database, Agentic AI ready | Open Source, Database |
The week’s signals
1. Agents are selling completed work
Ami AI targets customer acquisition, Clueso MCP turns inputs into editable videos, ProductBridge connects support to product feedback, and Naise AI focuses on marketing execution. The shared value is not “a smarter chatbot”; it is fewer manual handoffs between a goal and a deliverable. For a founder, start with a result that can be checked, then decide whether an agent is the right implementation.
2. Structured output and portable interfaces make agents usable
Jev’s official documentation describes typed decisions such as Choice and Score, and tells users to verify gateway pricing separately. Its SDK guide shows the product’s role clearly: make model judgment consumable by software. Superset Mobile tackles the human side, bringing workspaces, diffs, and pull-request review to an iPhone.
3. Permission and runtime safety are becoming the product
Solid and Sai move agents toward operating computers and accounts, while Arcjet focuses on prompt injection, tool authorization, sensitive-data redaction, and abuse prevention. Sider Omni Sidebar puts an agent beside existing Mac apps. The opportunity is real, but so is the warning: permission boundaries, audit trails, spending limits, and recovery may be more defensible than the automation button itself.
Focus products
#1 — Ami AI: customer acquisition as a measurable workflow
Product Hunt describes Ami as reading a website, selecting likely buyers, planning outbound, then running and monitoring campaigns. It appears to be a team-facing SaaS product, but this review found no reliable public plan or funding announcement. Its lesson is sharp: sell a measurable commercial outcome, not a generic “AI sales assistant.” Upvotes: 625 | Comments: 213
#2 — Clueso MCP: conversational video production
Clueso starts from an idea, deck, reference video, or recording and handles storyboards, scenes, voiceover, music, and editing while keeping the result editable. Its official pricing page confirms a free plan or trial without a credit card and offers custom plans beyond listed limits. Funding was not reliably public in this review. Upvotes: 539 | Comments: 160
#3 — Jev: decisions software can consume
Jev turns unstructured state into typed decisions instead of another long answer. Its hosted API model is aimed at routing, classification, risk checks, and automation. Public search produced third-party funding claims, but no primary announcement reliable enough to cite here. HN discussions include Introducing System One Models and Jev and Jev in 25 Lines of Python, with debate around latency, typed decisions, and paying for a managed service. Upvotes: 534 | Comments: 15
#4 — Superset Mobile: continuing work away from the desk
Superset Mobile lets users continue coding-agent work from an iPhone, inspect diffs, and handle pull requests. The product is bundled with Superset Pro; this report does not estimate an unverified price. Funding was not publicly confirmed. The useful business insight is that a mobile companion does not need to recreate an IDE: it can focus on status, review, and approval. Upvotes: 504 | Comments: 135
#5 — AINA: turning job-search blind spots into next steps
AINA uses conversation and a video avatar to help job seekers identify weaknesses and leave with a personalized plan. Pricing and funding were not reliably public in this review. Its product lesson is simple: a personal AI service becomes more useful when it ends in an action, not just encouragement. Upvotes: 497 | Comments: 197
#6–#8: three ways to make execution real
ProductBridge connects support, surveys, votes, and product feedback for customer-success teams. Solid gives agents computers, accounts, and budgets, which makes permissions and spending limits central adoption questions. Naise AI and MosMos also had no funding status that could be confirmed from reliable primary sources. Sai uses autonomous computers for screen-based work, which may feel more intuitive than API integrations but deserves extra caution around logins and payments. Their funding and complete pricing details were not reliably public in this review. Product Hunt counts were 474/55, 458/47, and 421/48 for upvotes/comments respectively.
Founder ideas to borrow
- Build a small support-feedback sorter that turns three scattered sources into the three product issues worth fixing today.
- Build an approval layer for agent actions, showing permissions, estimated spend, and recovery before a human clicks approve.
- Build a mobile handoff layer for solo founders: status, diffs, errors, and approvals, without rebuilding the desktop workspace.
Risk disclosure
Product Hunt votes are discovery signals, not proof of product-market fit or revenue. The launch data and counts above come from the prefetched JSON; pricing, funding, and performance cannot be inferred from votes. Most Top 10 funding statuses are therefore marked as not publicly confirmed rather than filled with weak search claims. Before testing a computer-using agent, choose a reversible task with no payment authority, narrow permissions, and an easy human review path.
Was this article helpful?
