Cycling through all six. Tap any point to stop.

Measured on six things

Permanent web hosting for AI agents, no human step

Security

HTTPS + HSTS, DMARC/SPF, signed version manifests, token-scoped ownership. But no CSP, no frame protection, no Permissions-Policy, and no published SOC 2/ISO or bug bounty found.

Screenshots

1 / 1

deed.page

About this app

deed.page is the publish step for AI-generated pages, reports and prototypes. Sites are permanent by default; an agent can set X-Ttl for pages that should expire. Launching now with a free plan anyone can try in one curl command. One PUT or MCP call publishes files to a permanent URL at a slug the agent picks, plus a token to update it later. No signup, no emailed code, no claim link, no 24-hour expiry.

Key features

1One PUT or POST publishes files to a permanent URL at an agent-chosen slug; response returns a Bearer token that owns the slug
2Native MCP server at /mcp with 26 tools, plus OpenAPI spec, llms.txt and markdown mirrors of every doc page
3Free tier with no signup: 25 sites × 25 MB, 60 publishes/hour, 1 custom domain, 1 webhook, cookieless analytics

Target customer profile

Drawn from the product itself, not from a survey.

Demographic

AI agent developers and automation engineers

Pain points

Struggle to find hosting that allows AI agents to publish autonomously without signup or email verification.

Primary needs

A simple API that enables one-request publishing and permanent URLs with ownership tokens.

Demographic

DevOps and platform teams in tech companies

Pain points

Need a lightweight static hosting solution for internal tools, prototypes, and reports without complex setup.

Primary needs

Fast, reliable hosting with version control, webhooks, and custom domains via API.

Demographic

Indie hackers and solo developers

Pain points

Want to quickly share demos, landing pages, or documentation without dealing with traditional hosting accounts.

Primary needs

Free, no-signup hosting with permanent URLs and easy updates.

Scout Review

Written by AI from measured evidence, scored out of 100.

deed.page is a focused, well-built answer to a real problem: letting AI agents publish static sites without a human step. The one-request publish flow, complete API docs, OpenAPI spec and 26-tool MCP server are unusually mature for a product this young, and the measured speed (98 mobile / 100 desktop) backs up the pitch. Design is clean with a couple of cramped spots, security is solid-but-unattested (8/12 headers, no CSP, no SOC 2), and accessibility is strong at 97/100 with three fixable fails. The comparison table against here.now is sharp positioning but poor accessibility. Growth is the strongest argument: the wedge is real and the incumbents are structurally slow to copy it. The main risk is that 'no signup, no claim link' is also a trust problem — there's no published security attestation to reassure the platform teams in the target list.

Measure by measure

Usability

One curl PUT to a live URL, full API reference with OpenAPI spec, llms.txt, markdown mirrors, and a 26-tool MCP server. Onboarding is literally the first request. Docs are unusually complete for a launch-stage product.

Design

Clean, confident developer-site aesthetic: big hero type, readable monospace code blocks with copy buttons, tidy numbered how-it-works cards. Comparison table and tiny hero stats line are the weak spots.

Speed

Lighthouse 98 mobile / 100 desktop, LCP 1.8s, CLS 0.017, TBT 61ms. A static marketing site that actually behaves like one — no heavy SPA tax, no layout jump.

Security

HTTPS + HSTS, DMARC/SPF, signed version manifests, token-scoped ownership. But no CSP, no frame protection, no Permissions-Policy, and no published SOC 2/ISO or bug bounty found.

Accessibility

Lighthouse 97/100 with three real fails: color-only link distinction, mismatched accessible names, and headerless cells in the big comparison table. No WCAG statement or i18n found.

Growth

Sharp wedge: agent-native publishing with no claim link, permanent by default, MCP built in. Real problem, specific customer, clear reason to pick it over here.now and Netlify. Early (domain ~weeks old) but the idea has room.

Improvement areas

Missing CSP, X-Frame-Options and Permissions-Policy; no published SOC 2, ISO or pen-test attestation
Comparison table lacks <th> headers and links rely on color alone — real WCAG fails on the main sales page
Hero stats line is too small and the FAQ has no search or category grouping
Lost-token recovery depends on an optional email; no recovery path without one

Design & Usability

deed.page looks like what it is: a developer tool that respects developers. The hero is bold, the code blocks are legible with working copy buttons, and the numbered how-it-works section explains the model in three cards. The comparison table against here.now is the design low point — full sentences crammed into cells, cramped on desktop, and the hero's '2,300+ sites' proof line is set so small it undercuts its own credibility. Usability is where it shines: the entire onboarding is one curl PUT, and the docs are a genuine API reference with an OpenAPI spec, llms.txt, markdown mirrors of every page, and a 26-tool MCP server. The upgrade path is documented down to the error response that hands your agent a checkout link. The only friction is human-facing: a long FAQ with no search and a DASH button that blends into the nav.

Speed & Security

Speed is the standout measured result: Lighthouse 98 on mobile and a perfect 100 on desktop, LCP 1.8s, CLS 0.017, TBT 61ms. For a marketing site that has to sell a hosting product, that's the right first impression — it loads like the thing it's promising. Security is competent but not exceptional. The probe confirms HTTPS enforced, HSTS, X-Content-Type-Options, Referrer-Policy, SPF and DMARC, and the docs describe Ed25519-signed version manifests plus token-scoped ownership. But CSP, frame protection and Permissions-Policy are all absent, and there's no published SOC 2, ISO, pen-test or bug bounty in the site content. For a product whose entire value proposition is 'hand us your files with no account,' the missing CSP and the lack of any third-party security attestation are the two things worth fixing next.

Accessibility & Growth

Accessibility lands at 79: Lighthouse scores 97/100, but three real fails remain — links distinguished by color alone, visible text labels that don't match accessible names, and a large comparison table whose cells have no headers. That last one matters because the table is the product's main sales argument; a screen reader user gets the pitch as an undifferentiated wall. There's no WCAG statement and no multi-language support in the site content. Growth is the most interesting dimension. The problem is real and specific: AI agents need to publish without a human in the loop, and the alternatives either expire in 24 hours (here.now) or demand an account (Netlify, Vercel). deed.page's wedge — permanent by default, agent-chosen slug, token as the deed, native MCP — is a genuine structural difference the incumbents can't easily copy without breaking their account model. The site claims 2,300+ sites published and 1,400+ in the last week, which is encouraging context but not proof of a market. The domain is weeks old. The idea has room; the execution of the wedge is what will decide whether it fills it.

Conclusion

If you're building agents that need to ship artifacts to a URL, deed.page is the cleanest option in its category right now — the one-request publish, the token-as-ownership model, and the MCP surface are all genuinely well done. If you're a platform team evaluating it for internal tools, the missing CSP and the absence of any SOC 2 or pen-test disclosure are the questions to ask before you route production artifacts through it. The product is early, the idea is sound, and the execution on the core loop is better than most launch-stage tools. Worth a curl; worth a harder look before it holds anything sensitive.

Suggestions for improvement

Add a Content-Security-Policy, X-Frame-Options and Permissions-Policy to close the remaining header gaps.
Publish a security page with pen-test results or a bug bounty — 'no signup' needs trust signals for platform teams.
Fix the comparison table: add <th> headers, and stop relying on color alone to distinguish links.
Enlarge the hero stats line and add search or category grouping to the FAQ to reduce scroll fatigue.

How we score

Alternatives comparison

Named competitors, point by point. Nobody paid to appear here or to be left out.

	deed.page	here.now	Netlify	Vercel
Human step required before a site is permanent	None — first publish is permanent by default	Claim link or emailed sign-in code	Account signup and configuration	Account signup and Git integration
Free tier limits	25 sites × 25 MB, no signup, 60 publishes/hour	10 GB / 500 sites after emailed code; anonymous expires in 24h	Free tier with account; build minutes and bandwidth caps	Hobby free tier with account; bandwidth and function limits
MCP / agent-native interface	Native MCP at /mcp, 26 tools, Streamable HTTP	MCP not currently advertised as supported	No native MCP; API and CLI	No native MCP; API and CLI
Paid entry price	Hobby $3/mo, Pro $15/mo	Hobby $4/mo, Developer $20/mo	Pro tier priced per seat/member	Pro $20/user/mo
Version history and rollback	Every plan: Free 10, Hobby 50, Pro unlimited; public v{n} permalinks	Account plans only; anonymous sites have none	Deploy history with instant rollback	Deployment history with instant rollback
Custom domains	Via REST API and MCP; Free 1, Hobby 3, Pro 20	Yes, including registrar	Yes, with automatic HTTPS	Yes, with automatic HTTPS