Product Feature Research 2026-04-28 · v0.1 · 5 competitors · 126 pages
Strategy

CreateOS wedge, the unassembled stack, and the moves to make next.

Reframe everything around "the only agent sandbox where the agent can earn." MPP Gateway + router.nodeops.network + marketplace is the unassembled stack — no competitor has it. The window before Cloudflare absorbs the agent-economy narrative is measured in months.

Wedge · Critical BD gap · Top 3 moves · 10 headline findings · Hockey-stick
The thesis

Your wedge is not prompt-to-app. Your wedge is the only agent sandbox where the agent can earn.

Bolt owns "prompt-to-app in 30s." Lovable owns "React+Tailwind+Supabase done beautifully." Fighting them on their axis means taking on an 18-month head start with a 20%-better product — and that's been proven to not work.

You have primitives none of them have assembled: MPP Gateway (on-chain agent payments, live), router.nodeops.network (payment-aware inference routing, live), and a creator-revenue-share marketplace. The combination is the category. Reframe everything — homepage, pitch deck, founder narrative, launch moment — around "the only agent sandbox where the agent can earn." That's the positioning that unlocks OpenAI Agents SDK inclusion, LangChain partnership, and enterprise agent deployments.

The unassembled stack
CapabilityCreateOSClosest peers
Agent sandbox✓E2B · Daytona · Modal · CF
Inference routing✓ routerCF AI Gateway
HTTP 402 payment rail✓ MPP— none
Marketplace + rev share✓— none
Deployment + DBs✓Vercel · Render · Railway
Critical BD gap — highest leverage in the next 30 days
CreateOS is not on OpenAI's Agents SDK default provider list.
The current list is Blaxel, Cloudflare, Daytona, E2B, Modal, Runloop, Vercel. Every AI agent builder in 2026 will read this SDK documentation. Named contact: Karan Sharma (OpenAI, Agents SDK product). Process: build reference integration implementing the sandbox provider interface (2-4 weeks engineering), publish case study of agent earning real money on MPP, submit. One BD win here = a year of content marketing. If OpenAI declines, open-source the integration — still delivers distribution via the SDK community.
01
Reframe positioning publicly + resolve x402 alignment
Homepage, pitch deck, founder narrative all lead with "the only agent sandbox where the agent can earn." Publish MPP ↔ x402 technical alignment doc — pick facilitator / extension / compatible, commit. Register with x402 Foundation as ecosystem participant, not rival protocol.
Week 1 · Owner: NK + marketing + MPP eng lead
02
Build + submit OpenAI Agents SDK integration + land first MPP case study
Implement the sandbox provider interface against CreateOS. Ship one reference customer with an agent earning real money on MPP — single most important piece of evidence that the positioning is real. Submit integration to Karan Sharma by end of Week 4.
Weeks 1-4 · Owner: Engineering lead + NK
03
Agent Economy Week — coordinated 5-day launch
Cloudflare-Agents-Week-style blitz: Day 1 positioning keynote. Day 2 customer cases (3+ live earning agents). Day 3 ecosystem announcements (LangChain, x402, Virtuals). Day 4 product (router.nodeops). Day 5 developer deep-dive. Target: 5% of CF Agents Week attention.
Month 2-3 · Owner: NK + marketing
The three-factor hockey-stick formula
(new technical primitive) × (narrow product decision) × (coordinated distribution event) = vertical curve
Your primitiveMPP + router.nodeops.network + marketplace with revenue share — assembled, live, differentiated. The combination is the category.
Your narrowing"Deploy an earning agent" keystone event, not "prompt-to-app in 60s." Use primitives only you have. Stop fighting Lovable on their axis.
Your eventAgent Economy Week in Month 2-3 — not Month 10 as the original playbook sequenced. Agent infrastructure category is consolidating real-time.
Every breakout in the reference set (Bolt, Lovable, Cursor, Replit, Base44, Hugging Face, v0) combined all three factors within a 1-3 month window. Railway is the only non-hockey-stick — they deliberately chose slow-compound. Any zero factor produces a linear curve, not a vertical one.
Overview Positioning & wedge 10 gaps Three-factor inflection BD targets 30-60-90 plan Risks

Headline findings — and what they mean for CreateOS

Ten observations across Replit + Anything + Fly + Hercules + Multica.
01
The landscape is four lanes — and three of them now have multiple players.
Builder lane (3): Replit + Anything + Hercules — converging on the same shell, fighting over makers. Runtime lane (1): Fly.io / Sprites. Orchestration lane (1): Multica (open-source agents-as-teammates). CreateOS: you're in a fifth lane — agent-economy infra (sandbox + router + MPP + marketplace) — that nobody else is in. Three competitors fighting for makers is structurally good for whoever picked a different lane.
02
Fly.io is the actual competitor — not Replit or Anything or Hercules.
Sprites = hardware-isolated sandboxes pitched for "agents that scale to millions of users." Same slot CreateOS wants on the OpenAI Agents SDK list. CreateOS: head-to-head is with Fly / E2B / Modal / Cloudflare. Three axes: isolation, per-agent spend bounds, and on-chain payments (MPP). Fly cannot match MPP without entering money movement. The 3 builders aren't competitors — they're potential customers if/when their generated apps want HTTP-402 endpoints.
03
Payments is the sharpest product-level difference (builder side).
Anything.com treats payment acceptance as a first-class product surface. Replit treats it as integration #N among 100+. CreateOS: MPP Gateway compresses this further — autonomous agent transacts on-chain. Neither competitor can replicate without entering money movement.
04
Autonomy is no longer a differentiator — it's how it's packaged.
Replit bundles 5 agent features into Core; Anything.com unbundles autonomy as Max SKU. CreateOS: offer an agent-specific pricing tier competitors structurally can't: per-agent spend caps, budget isolation, MPP earning dashboard.
05
Anything.com is quietly a dev tool.
Screenshots surfaced a full code editor, Figma import, conventional-commit labels, Secrets management — none marketed. CreateOS: don't repeat the mistake. Pick one demo stack (Next.js + Tailwind + PostgreSQL) and lead with it.
06
Security + SBOM is Replit's hardest moat.
Replit Core includes SBOM, CVE scanning, scan history — procurement artifacts. CreateOS: agent-era equivalent is per-agent spend caps + audit trails. Ship per-agent spend caps before MPP mainnet scale.
07
Named design systems are Anything's hidden moat.
Anything's agent snaps to curated presets ("High-Fidelity SaaS", "Ghost Structure"). CreateOS: analog is named agent presets — "research agent", "trading agent", "support agent" pre-wired to MPP + router.
08
The "Made with Replit" badge is a viral growth loop.
Every free-tier app ships with watermark. Clicks earn creator 50 credits. CreateOS: ship "Made with CreateOS" + a credit bribe. Or attribute via micro-signature on every MPP transaction. Turn every earning agent into distribution.
09
Multica is a partner, not a competitor — and the integration shape is already designed.
Multica = open-source agent orchestration (Linear-for-AI-coding-agents). 21.3k GitHub stars. Vendor-neutral by design — works with 10+ coding-agent CLIs. Their "Unified Runtimes" panel is built for pluggable cloud runtimes. CreateOS: ship a Multica-compatible cloud runtime adapter. Tasks dispatched from Multica's assignee picker land in CreateOS sandboxes; MPP earnings flow back as a metric on the runtime card. Add Multica as Tier 2 BD. Risk: if you don't, Modal / E2B / Daytona will become the default runtime for that 21.3k-star community first.
10
Three converged builder shells > "we're a better builder" pitch.
Replit, Anything.com, and now Hercules are all running essentially the same product (prompt → app → preview → publish, with built-in auth/db/payments). Hercules's "100k+ users" claim + Anything's monetization-stack moat + Replit's enterprise depth means the maker-builder lane is full. CreateOS: reinforces the "different lane" thesis. The convergence is the argument for the agent-economy positioning — every builder needs a runtime + payment layer below them, none of them want to build money movement themselves. You're under all three of them, not next to them.