[PetrefiedThunder]

Moves the agentmiddleware.dev homepage into this repo, per C.Lee's decision, so Vercel can build it with root directory site.

Commits:

1. Adds site/ copied byte for byte from archived agentmiddleware/agent-middleware-api at source commit 9289867d50c6cbc184db3efc5ef630378fb5f973 (via git archive). Verified: the site tree hash here equals the source tree hash (b921339b).
2. Applies the site/ changes from Homepage: hero receipt copy, example output, single install block, remove curl text agent-middleware-api#852 on top. The resulting site tree equals #852's head site tree (5a3c1055).

Not included: tests/test_site_agent_interface.py from #852. It imports the API backend (app.core.product_positioning, app.routers.well_known), which doesn't exist in this repo, so it can't run here. The site's own node --test suites are included.

Local checks: all 6 site/test_*.mjs suites pass with node --test. build_site.py needs the launch contact env values (PUBLIC_CONTACT_EMAIL, PUBLIC_DISPLAY_NAME) that Vercel provides, so it wasn't built locally.

No Vercel settings changed.

---

Note

Low Risk
Adds static marketing assets and a gated build pipeline; it does not modify gateway or auth code, though demo-key pages call the live API and production deploy depends on correct Vercel env vars.

Overview
Adds a new site/ tree so Vercel can build and serve www.agentmiddleware.dev from this repository (content moved from the archived API repo, plus stacked #852 site/ updates).

Build and deploy: build_site.py renders HTML from tokenized sources into dist/, inlines nav/footer, expands proof transcript consoles, generates FAQ JSON-LD from visible markup, and blocks production builds until PUBLIC_DISPLAY_NAME and PUBLIC_CONTACT_EMAIL are real (optional booking URL and Vercel analytics flag). build_docs.py turns docs-src/ Markdown into /docs/ pages. package.json points the static builder at the Python build.

Surfaces: Marketing funnel pages (home, proof, pilot, compare, demo-key, try), generated docs, branded 404, and machine pointers (/.well-known/agent.json, security.txt, llms*.txt). Shared client assets include strict-CSP-friendly a11y preload/panel, optional Vercel Web Analytics event hooks, UTM capture, lazy-loaded arcade “waiting room,” and demo-key.js calling POST /v1/demo/keys on the canonical API.

Not in this PR: API-backend test_site_agent_interface.py from #852 (depends on modules that live only on the API app).

^{Reviewed by Cursor Bugbot for commit 9fdf53a. Bugbot is set up for automated code reviews on this repo. Configure here.}

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-24T04:39:00.506591Z	8a3a417	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[coderabbitai]

Caution

Review failed

The pull request is closed.

Walkthrough

The change adds a static marketing site with generated documentation, public product pages, deployment settings, and machine-readable discovery files. It adds browser-based demo-key and pilot tools, accessibility controls, analytics, an arcade loader, and a WebGL wave renderer. It also publishes a receipt, trust-key snapshot, and transcript, with browser display and offline verification support.

Estimated code review effort: 5 (Critical) | ~120 minutes

Finishing Touches

📝 Generate docstrings

• Commit to this branch
• Create a new PR

🧪 Generate unit tests (beta)

• Commit to this branch
• Create a new PR

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

_{Comment @coderabbitai help to get the list of available commands.}

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 8a3a417c64

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Use a recorded receipt in the homepage example

The homepage presents rcpt_7Kq2mX9 as quickstart output and repeats it in the Call and Retry steps, but this is synthetic: the gateway generates rcpt- followed by 16 hexadecimal characters (gateway/app/services/receipts.py:443), and the published receipt has a different ID. This violates the real-data-only rule in AGENTS.md. Render an actual recorded exchange or use the existing published evidence instead of inventing receipt output.

AGENTS.md reference: AGENTS.md:L53-L53

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Run the documented local commands from the gateway directory

For readers cloning the linked repository, the instruction to run this command from the repository root fails with No rule to make target 'prove-trust-plane'. A repository-wide search finds the Makefile only under gateway/, where both prove-trust-plane and quickstart are defined; the docs/quickstart.md reference likewise needs the gateway/ prefix from the root. Update these instructions and their copies in llm.txt, llms-full.txt, and the agent manifest to use the gateway working directory. Locally, make --help confirms that -C DIRECTORY means “Change to DIRECTORY before doing anything,” so make -C gateway prove-trust-plane is suitable from the repository root.

Useful? React with 👍 / 👎.

[PetrefiedThunder]

Superseded by #200: a fresh branch so gitleaks' full-history scan passes; same final site/ tree.