- You want a scheduled, event-, anomaly-, or signal-triggered company loop that reads revenue, traffic, search, and site health each run.
- You want trust levels 1 to 5, hard spending caps, receipts, and rules learned from your approvals and rejections.
- You want a local, file-first harness of Markdown contracts and JSONL journals you can review, commit, and run through Codex or Claude Code.
win.sh alternative
win.sh runs a company loop. Soloop runs one move at a time.
win.sh is the closer fit when you want a scheduled or signal-triggered AI CEO loop with trust levels, hard caps, receipts, and a file-first local harness. Soloop is the closer fit when you want specialists that execute research, social, and code today, with each consequential move approved before it runs.
Decision shortcut
Which operating model fits?
- You need execution across research, social, and code now rather than connectors marked Soon.
- You want a specialist to do the work after approval, not only a loop that acts or asks.
- You want the approval surface to be one legible card per move rather than a rule system.
Evidence ledger
Soloop vs win.sh, dimension by dimension.
Read each row as an operating choice, not a feature score.
Starting point
A new company or a connected existing one, with budget and rules set up front.
An existing idea, URL, repo, deck, file, demo, or rough product.
Next move
Each run loads revenue, traffic, search, site health, goals, rules, recent decisions, and memory, then acts or asks.
The Plan Agent proposes one next-move card with a reason and success signal.
Approval model
Spend, outreach, publishing, and sensitive changes hit approval gates; trust levels, hard caps, and learned rules govern the rest.
Research, posts, code changes, credit spend, and deploys wait for founder approval.
Build and deploy
Build connectors such as GitHub are marked Soon; hosted tasks run in ephemeral sandboxes.
The Coding Agent changes founder-provided code in a sandbox and prepares a deployable result.
Growth and sales
Only Stripe, Plausible, and Google Search Console are marked live; Gmail, HubSpot, ads, and support connectors are marked Soon.
The Analyst researches demand; the Social Agent runs approved Twitter/X posts and replies today.
Signals
Records facts, decisions, learnings, and operating rules each run; approval and rejection behavior can become future rules.
Replies, users, task notes, code changes, and failures return to the workspace for the next call.
Ownership
An open local harness keeps contracts and journals in your files; complete hosted-workspace export, credential architecture, and log retention are not established.
Starts around founder-owned artifacts; no managed infrastructure transfer path is claimed.
Packaging
A user-selected $50 to $10,000 monthly operating budget with top-ups and a hard cap; platform margin and cost conversion are not published.
No total-cost comparison because win.sh does not publish how budget converts to work.
Operating loop
The interface determines where judgment sits.
win.sh
- 1Set budget and rules
- 2Trigger a run
- 3Check live signals
- 4Act or ask
- 5Record rules and learnings
Soloop
- 1Read current evidence
- 2Propose one move
- 3Approve
- 4Act
- 5Return the signal
Evidence boundary
What the sources establish—and what they do not.
- win.sh, built by Yuki Capital, positions itself as an AI CEO that runs a scheduled, event-, anomaly-, or signal-triggered company loop.
- Its integrations page marks only Stripe, Plausible, and Google Search Console live; the other 37 catalogued connectors are marked Soon.
- Docs describe an open local harness of Markdown loop contracts and JSONL state that can run through Codex or Claude Code.
- Pricing is a user-selected $50 to $10,000 monthly operating budget with a hard cap and receipts as of the evidence date.
- The current privacy policy still describes an AI text-rewriting platform and warns against confidential business data, which conflicts with the company-operations product.
- How the operating budget converts to platform cost or margin is not published.
- Complete hosted-workspace export, credential architecture, log retention, and model-training treatment are not established.
Source ledger
Check the primary material.
Links open the official source used for the statement. Product details can change after the evidence date.
AI CEO and company-loop positioning
File-first local harness and Codex/Claude Code execution
Three live versus 37 planned connectors
Operating-budget range, hard cap, top-ups, and receipts
Human oversight, ownership boundary, and ephemeral hosted tasks
Stale text-processing description and data warnings
FAQ
Questions founders ask before switching.
Is Soloop a win.sh alternative?
Yes. Both gate consequential actions behind approval and feed signals back into planning. win.sh runs a triggered company loop governed by trust levels, caps, and learned rules; Soloop proposes one move at a time and has research, Twitter/X, and coding specialists executing today.
Which win.sh integrations are live?
As of September 1, 2026, the official integrations page marks Stripe, Plausible, and Google Search Console live. The remaining 37 catalogued connectors, including GitHub, Gmail, PostHog, HubSpot, ads, support, accounting, custom API, and custom MCP, are marked Soon.
How much does win.sh cost?
win.sh asks you to choose a monthly operating budget between $50 and $10,000 with top-ups, a hard cap, approval controls, and receipts. Public pages do not explain how that budget converts to platform cost or margin, so total cost cannot be compared directly.
Make the next move reviewable
Bring the product you already have.
Soloop reads the current state, proposes one move, waits for approval, and returns the result.
Start with a projectThis is an independent comparison based on public information checked September 1, 2026. win.sh and related marks belong to their respective owners. Soloop is not affiliated with or endorsed by them.