The Solo-Operator Agent Stack

The one-person company is no longer a thought experiment in 2026. Anthropic CEO Dario Amodei put a 70-80%^[1][2] probability on the first billion-dollar one-person company by 2026 in May 2025^[1][2], and within 12 months that prediction looked conservative. Maor Shlomo built Base44 alone, reached 250,000 users^[2] in six months, and sold it to Wix for $80 million^[2] cash in June 2025; the product then hit $100M^[2] ARR in nine months under Wix^[2]. Ben Cera's Polsia hit $1M^[3] ARR in 30 days post-launch with 1,300+^[3][4] active companies on the platform, all run by AI agents^[3][4]. Pieter Levels operates a $3M+^[5] ARR portfolio (Nomad List $3M+^[5], PhotoAI $1.58M^[5], RemoteOK $1M+^[5]) on <$200/month^[5] infrastructure, with zero employees^[2][5]. The canon is no longer anomalies — it's a repeatable playbook with a documented stack.

This paper is an authority survey of that stack as it stood in early 2026: who's running which agents, what each layer costs, where the substitutional economics actually land, and the four founder wedges still open inside the architecture.

#Executive Summary

1. The one-person-company canon is now well-documented. Amodei's 70-80% prediction (Code with Claude conference, May 2025^[2]) named three categories: proprietary trading, developer tools, and businesses with automated customer service. By Q2 2026 at least three solo founders had crossed canonical thresholds: Maor Shlomo (Base44 → $80M Wix exit, $100M ARR post-acquisition in 9 months)^[2], Ben Cera (Polsia $1M ARR in 30 days, 1,300+ companies on platform, claims $3.5M ARR by Q1 2026)^[3][4], Pieter Levels (PhotoAI $1.58M ARR + Nomad List $3M+ + RemoteOK $1M+ + 70+ failed startups before)^[5]. The categorical claim from the OpenBooklet survey: 44% of profitable SaaS products are now run by a single founder, doubled since 2018^[5].

2. The dominant solo stack is four agents at $300-$700/month. Per the Foundra and SalesSheet operator surveys, most solo founders are running between four and seven agents, costing $300-$700/month total^[6][7]. The four canonical categories:

   • Coding. Claude Code Max 5x at $100/mo + Cursor Pro at $20/mo as the dominant pair^[8][9][10].

   • Workflow automation. Make at $9/mo or Zapier at $29.99/mo or n8n self-hosted at ~$5/mo or Lindy at $49.99/mo^{[11][12][13][14][15]}.

   • Support. Intercom Fin resolves ~70% of L1 tickets^[5], Crisp Pro at $25/mo, Zendesk AI.

   • Content/ops. Claude or ChatGPT at $20/mo, Buffer/Typefully at $0-15/mo, Perplexity Pro at $20/mo^[5].

3. The substitution math is 95-98% cost reduction. Per the AgentMarketCap and Athenic frameworks: a 6-8 person team costs £240,000-£320,000/year^[16] (or in USD: $280K engineering + $120K support + $90K marketing = ~$570K^[1]); a solo founder running an equivalent agent stack costs $3,000-$12,000/year^[1] (£2,400-£4,800)^[16]. The headline ratio is 95-98% savings^[16] for the operational mechanics — the founder's own time is now the binding constraint, not the headcount.

4. Cursor's $29.3B^[9] valuation reframes "developer tools" as a $30B+^[9] category. Cursor closed a $2.3 billion^[9] Series D at a $29.3 billion^[9] valuation in November 2025, after hitting $1 billion^[9] ARR faster than any SaaS company in recorded history^[9]. The Cursor data point that actually matters for solo operators: 35%^[9] of all merged pull requests at Cursor are now created by autonomous AI agents (not "with AI assistance" — autonomously authored and submitted)^[9]. Agent users now outnumber Tab users 2:1^[9] — an inversion of the March 2025 ratio of 2.5:1^[9] in the other direction. Agent usage at Cursor grew 15×^[9] year-over-year through 2025^[9].

5. Claude Code is the outer-loop agent; Cursor is the inner-loop agent; most solo founders run both. The 500k.io comparison synthesis: 62%^[10] of solo founders running an AI-assisted workflow in 2026 use both tools side by side^[10]. The pricing combination is $120/month^[10] ($20^[10] Cursor Pro + $100^[10] Claude Code Max 5x). Claude Code has accumulated 5.2 million^[9] VS Code installs and writes an estimated 135,000^[9] GitHub commits per day. OpenAI's Codex CLI reached 2 million^[9] weekly active users by March 2026 — five months after launch^[9]. Replit closed a $400M^[17] Series D at a $9 billion^[17] valuation in 2026 (tripling from $3B^[17] six months earlier).

6. Workflow-automation platforms diverged on pricing axis. Per the Effloow and Dev.to comparisons: Zapier 750^[11] tasks at $19.99/month^[11] vs Make 10,000^[11] operations at $9/month^[11] = 13×^[11] more operations per dollar on Make. n8n self-hosted is ~$5/month^[11] for VPS infrastructure with unlimited executions, vs €720/year^[14] for n8n Cloud Pro at 10,000^[14] executions. Lindy's $49.99/month^[15] for 5,000^[15] credits is the AI-agent-builder positioning. The structural decision: Zapier for non-technical breadth, Make for high-volume cost efficiency, n8n for data sovereignty + technical customization, Lindy for natural-language AI-agent workflows^[12][11].

7. The Polsia / Athenic / Agent0 frameworks are templates, not products. Polsia ($49/month^[4] per customer, uses Claude Opus 4.6 as primary reasoning model^[4], AI CEO agent wakes nightly to evaluate the business, executes tasks, and sends a morning summary^[4]) is a hosted application of the framework. Athenic's 10-agent framework (Content Engine, Community Orchestrator, Research Analyst, SEO Optimiser, Email Nurture, Data Dashboard, Customer Support Bot, Outbound SDR, Quality Control, Integration Hub) is a how-to template^[16] documenting £1.8M^[16] ARR with 2 people and the 87%^[16] support-ticket AI resolution rate. Agent0 ($129^[18]-$269^[18] one-time, Obsidian vault + Claude Code framework^[18], 7 agents — Cortex/Loom/Radar/Hippocampus/Signal/Sentinel/Axon, 83^[18] skills across 10 categories) is the open-source-leaning framework variant.

8. The investor framing shift is now visible at the seed stage. Per the Foundra survey: instead of "when's your first hire?" investors are asking "which tasks have you automated?" and "what does your agent stack cost?"^[6]. The implicit cap-table math: a founder showing six healthy agents handling 60%^[6] of weekly operations looks more capital-efficient than one with three early hires burning $30K/month^[6] combined. Solo founders crossing $20K^[6] MRR are increasingly delaying their first hire until $50K^[6] or $80K^[6] MRR, and when they do hire, the first hire is a senior specialist, not a junior generalist^[6] — because a junior generalist's work is exactly what the AI stack already does for $100/month^[6].

#The Substitution Math

The single most-cited substitution table comes from AgentMarketCap's April 2026 analysis^[1]:

That's a 97%^[1] cost reduction for the operational layers. The Athenic framework's parallel UK accounting^[16]:

The match between independent operator estimates is the structurally important signal: two operators on different continents, working with different teams, both arrive at 95-98%^[1][16] cost reduction for the operational mechanics.

The OpenBooklet research adds the canonical ARR-band cost structure^[5]:

#Layer A: Coding Agents

The coding-agent layer is where the largest capital is being deployed. Cursor's Series D^[9] at $29.3B^[9] in November 2025^[9] is the highest valuation ever assigned to a developer tool^[9]; it's preceded by Series C at $9.9B^[9] (June 2025) and Series B at $2.6B^[9] (December 2024). Anthropic's Claude Code is the runner-up category leader on raw usage: 5.2 million^[9] VS Code installs and ~135,000^[9] GitHub commits per day. OpenAI's Codex CLI hit 2 million^[9] WAU five months after launch^[9]. Replit Agent 3 sits inside a $9B^[17] Series D platform with built-in deployment^[17]. Devin at $500+^[19]/month per agent is the premium async-task option^[19].

The pricing matrix that solo founders face^{[8][9][10][19]}:

The 500k.io guidance for the $500K-bound solo founder^[10]:

• Pick Cursor if your work is in-IDE flow — typing code, refactoring on the spot, fast tab-complete, multi-provider routing flexibility, solo-developer or small-team workload.
• Pick Claude Code if your work is agentic — multi-file refactors, long-running content/ops pipelines, repo-wide automation, MCP-server-heavy workflows.
• Run both if your week has both inner-loop coding and outer-loop ops/content. The combined $120/month bill is below the "non-issue" threshold for any post-revenue solo founder^[10].

The structural shift visible inside Cursor's own data is the most important read on where the trajectory points: agents now outnumber Tab users 2:1^[9], 35%^[9] of merged PRs are agent-authored, and Cursor's Cursor 3 launch (April 2026)^[20] explicitly repositions the IDE as a control layer — a place where developers assign tasks to agent swarms, review outputs, and set quality standards, rather than a place where developers write code themselves^[9]. The prompt box is gradually becoming more important than the syntax-highlighted file buffer.

#Layer B: Workflow Automation

The workflow-automation layer ties the rest of the stack together. The four-platform decision space converges around a clear axis^{[11][12][13][14][15]}:

The structural cost claim from the Effloow and Dev.to comparisons^[11][12]:

Make provides 13× more operations per dollar at comparable pricing tiers — 10,000 operations for $9 versus Zapier's 750 tasks at $19.99.

The structural sovereignty claim from the n8n side^[11][14]:

Organizations processing 100,000+ monthly operations find n8n's self-hosted deployment costs substantially lower than Zapier's enterprise-tier pricing. AI-heavy workflows benefit from n8n's advanced LangChain integration unavailable in simpler platforms.

The new entrant is Lindy ($49.99/mo Pro tier, 5,000 credits, 4,000+ integrations including Slack, Gmail, HubSpot, Salesforce + voice agents)^[15][21]. Lindy's positioning is fundamentally different from Zapier/Make/n8n: instead of building trigger-action workflows, the user creates "Lindies" — AI agents that handle tasks using natural-language instructions^[11]. Lindy is the AI-native heir to the workflow-automation category; Zapier and Make are the rules-based incumbents now scrambling to add agentic features (Zapier added MCP support and dedicated AI Agents in 2026^[12]; Make added agent capabilities; Relay.app explicitly merges automation with human-review checkpoints^[15]).

#Layer C: Support, Content, Infrastructure

The remaining stack layers are commoditized in 2026^[5]:

Support agents:

• Crisp Free at $0 (basic AI), Crisp Pro at $25/month for 20+ daily tickets
• Intercom Fin Essential at $29^[22]/seat/month — resolves ~70%^[22] of L1 tickets autonomously
• Zendesk AI agent layer
• Plain (modern alternative)

Content & marketing:

• Claude or ChatGPT at $20/month (copywriting, strategy, content drafts)
• Perplexity Pro at $20/month (AI research and content sourcing)
• Buffer or Typefully at $0-$15/month (social-media scheduling)

Coding-adjacent:

• v0.dev at $20/month (AI frontend generation with Vercel integration)
• GitHub Copilot at $10/month (inline code completion)

Infrastructure:

• Supabase at $0-$25/month (database + auth + storage)
• Vercel free tier for hosting

This layer is where the "free tiers everywhere" point compounds: a pre-revenue founder can stand up a complete operational stack at $0-$60/month total^[5].

#The Polsia / Athenic / Agent0 Operating Frameworks

Three distinct framework variants are now public:

Polsia (Ben Cera, San Francisco, Columbia engineering alumnus + ex-Cloud Kitchens GM)^[4]: A hosted SaaS at $49/month^[4] per customer. The product runs autonomous AI agents that "run entire companies": strategic planning, code-and-deploy, marketing/sales (cold email + Meta Ads + social), inbox management, even VC negotiations. Claude Opus 4.6 as primary reasoning model^[4]. The "AI CEO agent wakes up every night, evaluates the business, executes tasks, and sends a morning summary"^[5]. Cera's stated thesis: "80%^[3] AI, 20%^[3] taste" — agents handle the operational mechanics, humans hold judgment + product direction^[3]. Polsia hit $1M^[3] ARR in 30 days, manages 1,300+^[3] companies, and a self-reported $3.5M^[3] ARR by March 2026 (figures unverified, broadcast on a live dashboard at polsia.com/live)^[3].

Athenic 10-agent framework^[16]: A how-to template documenting £1.8M ARR with 2 people. The 10 agents:

1. The Content Engine — Claude 3.5 Sonnet + Custom GPTs + Athenic; 20 min/day human review
2. The Community Orchestrator — Zapier + Make + Athenic; 30 min/day human review
3. The Research Analyst — Perplexity AI + GPT-4 + custom scrapers; 15 min/week human review
4. The SEO Optimiser — Ahrefs API + AI + custom scripts; 1 hour/week human review
5. The Email Nurture System — Customer.io + AI personalization; 2 hours/month human review
6. The Data Dashboard — Retool + Athenic + custom Postgres; 10 min/week human review
7. The Customer Support Bot — Intercom AI + custom knowledge base; 45 min/day (down from 4 hours/day)
8. The Outbound SDR — Apollo + Clay + AI personalization; 1 hour/day human review
9. The Quality Control System — Custom GPT-4 fine-tune; 30 min/day approval
10. The Integration Hub — Athenic / MCP-based orchestration; 2 hours/week human review

The case study^[16]: a SaaS platform for freelance designers with one founder + one part-time developer running all 10 agents — £1.8M^[16] ARR, 3,400^[16] customers, 87%^[16] support tickets resolved by AI (Tier 1), 450^[16] pieces of content published, 12 sales deals at avg £35K contract value.

Agent0 (Maciek Marchlewski, alpha launched April 2026)^[18]: An Obsidian vault + Claude Code framework — explicitly local-first, Markdown on machine, syncs via GitHub. 7 specialized agents (Cortex orchestrates, Loom writes, Radar finds opportunities, Hippocampus verifies claims, Signal distributes, Sentinel monitors, Axon reviews code) with 83 skills across 10 categories. Human-in-the-middle governance: every action is auto-execute or requires-approval; a rejection log teaches the system. $129 Starter (one-time, self-install) / $269 Pro (one-time, pre-wired) / $2K-$10K Custom pricing^[18]. The TAM logic: ~28M non-employer businesses in the US^[18] + 300M+ self-employed globally^[18] + Obsidian's >1M MAU + Notion's $10B+ valuation = a defensible founder-side wedge.

#The Investor-Framing Shift

The most under-discussed implication of the agent-stack era is the cap-table math change at the seed stage. Per the Foundra April 2026 survey^[6]:

Investors are asking different questions^[6]. Instead of "when's your first hire?" they're asking "which tasks have you automated?" and "what does your agent stack cost?" A founder who can show six healthy agents handling 60%^[6] of weekly operations looks more capital-efficient than one with three early hires burning $30K^[6] a month. The cap table math has changed. So has the conversation in pitch meetings.

The downstream behavioral change^[6][16]:

• Solo founders crossing $20K^[6] MRR are now running agent stacks instead of hiring.
• Many won't make their first hire until $50K-$80K^[6] MRR.
• When they do hire, the hire is a senior specialist, not a junior generalist^[6] — because a junior generalist does what the AI stack already does for $100^[6]/month, with negative context-overhead.
• Capital efficiency has become a board metric, not a back-burner concern.

The "SaaSpocalypse" event of February 2026^[6] — roughly $285 billion^[6] in market value evaporating from software stocks in a single trading session^[6] as a wave of AI agent releases let one person do the work of five separate SaaS subscriptions — is the supply-side analog of the cap-table math change. The category most exposed to direct substitution is SaaS bundles whose primary value proposition is "we save your team time on a recurring task".

#The Founder Wedges

Four wedges are still open inside the architecture:

#1. Vertical agent-stack templates per ICP

The Athenic 10-agent framework and the Agent0 7-agent framework are horizontal — they work across industries. The wedge is vertical: solo-founder agent stacks specifically tuned to a named ICP (e-commerce DTC operators, fractional consultants, content creators, real-estate investors, indie SaaS founders). The deliverable: a curated stack with opinionated defaults, prompts, integration manifests, and skills, packaged as a one-time purchase ($129-$2K) or monthly subscription ($39-$200/mo for Knolli-equivalent^[23]) per ICP. The ChatFin / Knolli pattern^[23]+ Agent0's per-ICP variant.

#2. Agent-skill marketplaces for solo operators

Agent0 names this as a "future recurring layer — hosted agent runtime, shared vault for small teams, skill marketplace"^[18]. The pattern: a marketplace where solo operators publish (and sell) skills that plug into their fellow operators' agent vaults — outbound-email-writer skill, churn-detection skill, freelance-invoice-collection skill. The Cursor MCP marketplace + Claude Skills directory + Replit's bundles all hint at this — but the explicit fractional-CFO-style "skill = professional service productized" version is the founder wedge.

#3. Audit-trail / governance compatible with §2550.404a-6

Per the trump-eo-14330-401k paper^[24], the DOL's March 31, 2026 NPRM creates a process-based safe harbor requiring documented evidence of objective consideration on six factors. As solo operators sell into 401(k)-sponsoring small businesses (the natural evolution of the founder-velocity wedge from $20K-$80K MRR into the SMB market), the audit-trail requirement becomes a defensible product. Not just "log who did what when" — but "produce a §2550.404a-6-compatible record set per recommendation per quarter". The vendor that builds this layer first inherits the same compliance moat that Datadog inherited from "log shipping" in the previous decade.

#4. The founder-velocity research practice itself

The fractional-CFO practice has had 20+ years to develop its repeatable consulting layer. The fractional-CTO practice has had 5+ years. The fractional agent-stack-architect / founder-velocity practice is 2 years old and growing. The wedge is the practice itself: a fractional-CFO-equivalent who specializes in standing up the 4-7 agent stack for a solo founder, tuning it over the first 6 months, and disengaging after the founder reaches $50K-$80K MRR^[6][16]. Pricing logic mirrors the fractional-CFO market^[24]: $5K-$15K/month retainer for the first 6 months, transitioning to project-based for the second 6 months. TAM: 2,500 active fractional CFO providers in the US^[24] is the comp; founder-velocity practitioners are at ~10% of that count today and growing.

#What This Paper Does Not Cover

This paper is an authority survey of the solo-operator agent stack as it stands in early 2026. It does not cover: (a) the headcount transition mechanics when a solo founder makes their first hire (fast follow paper), (b) the enterprise agent-stack architecture (different supply chain, different procurement dynamics — the Vercel BotID / Cloudflare verified-bot identity layer is a fast follow), (c) the specific ROI economics of the Polsia hosted-business model (waiting on more public data; current $3.5M ARR figures are self-reported), (d) the SaaSpocalypse aftermath and which SaaS categories were structurally hit hardest, (e) the legal-and-regulatory framework for AI-agent-authored work product (which solo-operator content qualifies as copyrightable under the Thaler v. Perlmutter precedent^[25]).

#References