The category is real and already contested: display.dev is executing well, Shopify proved the demand internally, and the LLM vendors ship native publish. None of them is open, self-hostable, and vendor-neutral at the same time. That intersection is the whole opportunity.
Position is directional, not measured. Gist/Pages and Netlify are "open-ish / neutral" on infra but are general web hosting with no artifact loop: no comments, no agent read-back, no versions-as-feedback.
Feature reality
display.dev matches us on the loop and is the team to beat. The bottom three rows are where it (and everyone) cannot follow without abandoning their model.
| Capability | Atrium | display.dev | Handoff | Claude / GPT | Netlify / Vercel | GitHub Gist |
|---|---|---|---|---|---|---|
| Permanent URL for HTML + MD | ● | ● | ● | ◐ | ● | ● |
| Inline comments | ● | ● | ○ | ○ | ○ | ◐ |
| Revision history | ● | ● | ● | ○ | ◐ | ● |
| Agent read-back (MCP) | ● | ● | ◐ | ○ | ○ | ○ |
| Works across all LLMs | ● | ● | ◐ | ○ | ● | ● |
| SSO-gated private artifacts | ● | ● | ◐ | ○ | ◐ | ○ |
| Built for throwaway artifacts | ● | ● | ● | ● | ○ | ◐ |
| Open source | ● | ○ | ○ | ○ | ○ | ○ |
| Self-hostable in your VPC | ● | ○ | ○ | ○ | ○ | ◐ |
| Open protocol (not a silo) | ● | ○ | ○ | ○ | ○ | ○ |
● full · ◐ partial · ○ none. Competitor positions are directional analysis, not vendor-confirmed.
Know thy rivals
"Agent-native publishing engine." dsp publish, MCP, SSO-gated URLs, inline comments with an agent republish loop (short_id / base_version), flat pricing, unlimited viewers. Sharp positioning: "the best agent changes, the home doesn't."
First mover on the exact loop, clean DX, already owns the "any agent, any model" message. Aggressive value framing (7-40x cheaper than Vercel/GitBook for gated sharing).
It is closed and cloud-only. No source, no self-host, enterprise starts at $499+/mo and artifacts still live on their servers. Regulated buyers cannot adopt it. We take the same loop and make it a commons they can run.
Internal hosting platform. Drop a folder, get a gated URL. GCS + wildcard NGINX + identity proxy, plus backend APIs (DB, AI, realtime). 50,000+ sites, half the company publishing, ~$200/mo.
That demand is enormous and that the architecture is cheap and simple. Validates both our market and our technical approach at scale.
It is not a product: Shopify-only, not open, no comments or version loop, deliberately no permissions model. It is the in-house thing every other company now wants and cannot build. We are that, packaged and open.
Built-in "Publish" that mints a public URL on the vendor's domain. Zero friction inside that one tool.
Default, free, in the box. Most casual sharing will start here. Massive distribution.
Vendor-branded, search-indexed, no auth, no comments, no versions, locked to one model. Useless the moment work is internal, multi-model, or needs iteration. They will not build cross-vendor neutrality or self-host: it is against their interest. That is structurally our lane.
The thesis
The same code runs on our cloud or inside the customer's VPC on standard tech. For banks, healthcare, gov, and any security-conscious team, "the artifacts never leave our network" is the unlock display.dev cannot offer.
An open Artifact Publish Protocol any LLM tool implements. As models churn, the home and its comment/version graph persist. The LLM vendors cannot be neutral about each other; we can.
Open source means security teams read the code, not a SOC 2 PDF. Lower the barrier to "yes" for the exact buyers who generate the most sensitive artifacts.
Clear-eyed
If Anthropic or OpenAI open-sources a cross-model artifact host. Counter: unlikely for them to be neutral about rivals or to ship self-host; if a standard emerges we adopt it and compete on the best open implementation + hosting.
They are ahead on product and messaging. Counter: we do not win on the loop, we win on open + self-host + protocol. Lead with the enterprise/regulated wedge they structurally cannot serve, and move fast in OSS where community compounds.
"Just a file host" race to the bottom. Counter: the moat is the comment + version graph and the protocol, not storage. Hosting is the loss-leader; the loop and the network of integrations are the value.
OSS that self-hosts may never pay. Counter: sell managed hosting + enterprise support/SSO/SLA (the Sentry/GitLab/Cal.com model). Keep the core genuinely good so adoption is wide.
Untrusted code is the attack surface. Counter: origin isolation + sandbox + proxy-gated auth from day one (see Technical). Treat it as the core engineering problem, not an afterthought.
Maybe artifacts stay throwaway. Counter: Quick's 50k sites and display.dev's traction say otherwise; the artifact is becoming the durable unit of knowledge work.
Where to push
Ship the Claude Code / Codex skill + CLI + a generous hosted free tier. Win the bottom-up loop display.dev and Handoff are fighting for: developer publishes an artifact, teammate comments, it spreads inside the org. OSS credibility is the acquisition channel.
When bottom-up usage hits a security review, the answer is already "self-host it in your VPC, here is the Helm chart, here is the source." That is the conversation display.dev cannot have. Convert on support + SSO + SLA.
The LLMs commoditize generation; the durable layer is where the output lives, who can comment on it, and how it changed over time. Own that as an open, neutral, self-hostable commons, and you become the default home no single model vendor is positioned to be.