Soloop Review (2026): Pricing, Features & Honest Verdict

Soloop is an AI agent workspace for solo founders that picks your next move, shows you a card explaining it, and runs the work only after you approve. The approval discipline is excellent. The execution bench behind it is thinner than the pitch. Best for: solo founders who already shipped something. Price: from $39/mo (free plan: yes). Rating: 6.5/10.

What is Soloop?

Soloop calls itself an “approval-first Agent OS for solo founders”. In practice it is a workspace where you dump the messy thing you already built, a repo, a landing page, a deck, a half-finished demo, and an AI agent labelled CEO reads it and proposes exactly one next move. You get a card. The card names the objective, the exact action, the target, the credit cost and the stop condition. Nothing runs until you approve that card.

We found it at the top of Product Hunt on 7 August 2026, launched by Wenhao Yu, who writes as Arthur Yu and is based in Singapore. His founder note is about watching demos die at Baidu because the teams could not work out what to do next. That is the problem Soloop is aiming at, and the framing is unusually specific: it starts after the first thing exists. The product docs on the site are stamped “current behavior, July 10, 2026”, which tells you how young this is. Worth flagging one thing up front: the site reads as a one-person project, while the Product Hunt listing names four people including a CTO and a CFO. Those two pictures do not match.

What Are Soloop’s Key Features?

The approval card

This is the whole product, and it is the best thing about it. Every action that posts publicly, spends credits, runs research, changes code or deploys generates a card first. You read the objective, the exact action, the target, the credit cost and the stop condition, then approve, request changes, or cancel. Coding cards add acceptance criteria so you know what “done” is supposed to look like before the agent starts. Most agent products bury this behind a confirmation dialog. Soloop makes the card the unit of work.

A named specialist bench

Work routes from the CEO agent to one of four specialists: CTO for product and code, CMO for acquisition, Analyst for competitor and market research, CFO for pricing and cost questions. The routing itself is honest. The CEO explains why it picked a specialist and does not run external actions itself. But read the docs carefully and two of those five roles do nothing but talk. CFO is described as “conversation today” and explicitly does not move money. CEO never executes. So the executing bench is three, not five.

Sandboxed coding with a real deploy

The CTO agent is the one with teeth. Point it at a new project, a GitHub repo, or a ZIP up to 50MB, and approved jobs run in an isolated sandbox that can return a checked change or push a Vercel deployment. Uploads are capped at 10MB per document for TXT, Markdown, PDF and DOCX, 3MB per image, five files per message, and 50,000 extracted characters per file in a chat message. Those are modest but published limits, which beats the usual silence.

A durable record instead of a chat log

Chat is treated as a view, not the source of truth. Four records persist: the Product Doc, uploaded files, Tasks (execution output, progress logs, failures and credit use) and Inbox (monitored replies). When you ask about earlier work you are told to name the file, task ID or Doc field so the CEO agent loads the right record. That is a deliberate design choice against the “scroll up and hope” pattern, and it is the right one.

What Can Soloop Actually Execute Today?

The launch copy promises an AI CMO that finds users and sells. The docs are more careful, and you should read them before subscribing.

Analyst research is live. Approved coding is live. X and Twitter are live once you connect OAuth, and the agent can draft, publish approved posts, and collect replies. Cold email is in beta and requires a connected Gmail account, one recipient at a time. Reddit is manual: Soloop drafts the post and opens a prefilled Reddit flow, you press publish. Ads are plan-only. Soloop will discuss ads and put them in a plan but does not buy or manage anything.

So the honest version is one fully automated public channel, one beta channel, one assisted channel and one that is just a conversation. For a product whose pitch is delegating acquisition, that gap matters. It is not dishonest, because the docs spell it out clearly and even say channels stay off the page “until there is runtime behind them”. But nobody reads docs before they read a homepage.

How Much Does Soloop Cost?

Soloop starts at $39 a month on the Builder plan, and there is a real free tier. Free gives you 15 credits a month, core planning and chat, and the same review-before-use gate. Builder at $39 gets 200 credits a month plus research, coding and growth workflows. Operator at $99 gets 600 credits and adds recurring operating runs plus a monitoring and approval queue. Team at $199 starts at 1,500 credits and steps up through 3,000, 4,500, 6,000 and 7,500 for teams and agencies running several projects. Annual billing takes 15% off.

Against the category that is reasonable on the surface. Lindy runs $49, $99 and $199 for its agent tiers, so Soloop’s ladder sits slightly below it at every rung. Manus prices on credits too. The problem is not the number on the card. It is what the number buys.

What Is a Soloop Credit Actually Worth?

Nowhere on the pricing page, the docs or the about page does Soloop publish what a credit costs you in work. There is no rate card. No “a research run is roughly X credits”. No worked example. The pricing page says only that monthly credits “fund work across planning, research, coding, and growth” and that one balance follows whatever you choose to run.

That is a real problem for the exact buyer this is aimed at. A solo founder deciding between $39 and $99 needs to know whether 200 credits is forty research runs or four. You find out by approving a card, because the card shows the cost, which means you find out after you have already paid for the month. The free tier’s 15 credits is presumably meant to answer this, but 15 of an undefined unit answers nothing.

Fix that and the pricing section becomes fine. Until then, budget for the possibility that Builder is a two-week plan rather than a monthly one, and treat the free tier as a metering experiment rather than a trial.

Who is Soloop Best For?

Use Soloop if you are a solo founder with something already built and the specific problem you have is deciding what to do next. The approval-card discipline suits people who want agents to do work but do not trust them unsupervised, and the Analyst and CTO agents are the two that actually execute. If your bottleneck is “I have a repo and a landing page and no idea which of eight things to try”, this is aimed squarely at you.

Skip Soloop if you need acquisition automated across channels, because today that means X and a Gmail beta. Skip it if you have no product yet, since the whole model assumes existing context to read. Skip it if you need to forecast spend before committing, because credits are undefined. And skip it if key-person risk bothers you: the about page describes a single builder, the launch names four people, and neither version is a company you would bet a critical workflow on in year one.

Best Soloop Alternatives

Lindy is the closest commercial comparison. It builds AI agents that run real workflows across email, calendars, CRMs and hundreds of integrations, with tiers at $49, $99 and $199 a month. Lindy has far more connected channels and a much longer track record. It has nothing like Soloop’s single-next-move framing, so you are trading direction for breadth.

Manus is the general-purpose autonomous agent people reach for when they want a task run end to end rather than a plan proposed. It also sells on credits. It will happily go further without asking, which is either the feature you wanted or exactly the thing Soloop’s approval gate exists to prevent.

Rolling your own is more viable here than in most categories. Claude Code plus a scheduler plus the X API gets you the coding agent and the posting agent, and it costs whatever your model usage costs. What you lose is the CEO layer choosing the next move, which is the part Soloop is really selling. If you already know what to do next, you do not need to buy this.

Final Verdict: Is Soloop Worth It?

Soloop has one strong idea and executes it well. Making the approval card the unit of work, with cost and stop condition visible before anything runs, is better agent design than most funded products manage. The durable-record model beats chat scrollback. The docs are unusually honest about what is live, what is beta and what is just conversation, which is rare enough to be worth saying out loud.

What holds it back is that the bench is thinner than the pitch and the pricing is unpriceable. Two of five agents cannot execute anything. Acquisition means one live channel. And a founder cannot work out what $39 buys without subscribing first. At 6.5 out of 10, this is worth the free tier and worth a month of Builder if the “what next” problem is the one keeping you up. It is not yet worth $199. Watch it: if the credit rate gets published and Reddit and email move out of manual and beta, this gets considerably more interesting.