Skip to content

Developer FAQ

Short, direct answers to the questions developers ask most. For symptom-driven fixes see Troubleshooting; for the full picture start with What is Sill.

Sill is agent-governance infrastructure that sits in front of a merchant’s existing commerce stack and answers three questions about every AI agent that interacts with it: identity (who is the agent), intent (what it is authorized to do), and proof (a signed, exportable record of what happened). Sill never custodies funds.

No. Discovery installs with a one-line script tag — no SDK, no package to vendor, no Sill code in your bundle. The signature-verification path is deliberately standard-only: any reputable RFC 8785 and ed25519 library works.

Live at the boundary today: A2A, AP2, and MCP. ACP, UCP, and x402 are on the roadmap, not enabled. See Protocols for the current status table.

The canonical embed snippet is:

<!-- paste before </body> -->
<script async src="https://cdn.sill.so/embed.js"
data-site-key="sk_…"
data-proof-token="pf_…"></script>

The data-site-key is the public site identifier (see below). The data-proof-token is a per-site, opaque ownership proof: the embed runtime does not read or transmit it. It exists only so Sill’s origin can server-fetch your published HTML and confirm the snippet (and therefore your control of the domain) is present. Both attributes are public-by-design once the snippet ships.

No. The site key is a public credential — it ships in every visitor’s page source, the same way a Stripe publishable key does. Seeing it in your HTML is not a leak. It identifies your site; it does not authorize sensitive actions.

No. Discovery is free and unlimited — identify agent traffic, publish your agent card, MCP server, and ARD catalog, all independently verifiable. Sill does not publish other pricing yet.

How do I verify that a Sill signature is genuine?

Section titled “How do I verify that a Sill signature is genuine?”

Fetch the public JWKS, canonicalize the manifest with RFC 8785 (JCS), and verify the detached ed25519 JWS against the matching key — using only off-the-shelf tooling, no Sill code. Full recipe with a JavaScript sketch and a negative-control test: Verify a signature.

No. Sill runs as managed infrastructure (the mandate/discovery engine at the edge, with audit and registry at the origin). You integrate by installing the embed and, for the transactional path, connecting your own processor.

No. Sill handles only opaque processor tokens (e.g. Stripe pm_* / tok_*); raw card numbers never enter any Sill system. This is an architectural property enforced by a CI gate — it is not a PCI certification. See Security.

US-East today (Phase 1). Sill does not publish latency or uptime figures.

Yes. Rate-limited requests return outcome: rate_limited with HTTP 429 and a retry_after_seconds value to back off against. The MCP server additionally enforces a per-session burst limit.

What’s the difference between Discovery and Transactional?

Section titled “What’s the difference between Discovery and Transactional?”

Discovery (free, live) identifies and logs which agents touch your site and publishes agent-readable manifests — read-only, no payment authorization. Transactional (Phase 2, early access) adds signed mandates, the policy engine, human-in-the-loop escalation, and payment authorization on your processor. Transactional is validated in limited production on the live Stripe rail — not scaled multi-merchant volume.

How do AI agents discover what my site offers?

Section titled “How do AI agents discover what my site offers?”

Through three signed, agent-readable surfaces Sill publishes per site: the A2A agent card, the MCP server (tools/list), and the signed ARD ai-catalog.json. All are verifiable against the public JWKS.

It’s held for human review (outcome: escalated, HTTP 202) — no charge occurs until a reviewer approves it. Each merchant decides which actions run unattended and which require review; the default policy ships conservative.

Can I use Sill’s framework mappings in my own audit?

Section titled “Can I use Sill’s framework mappings in my own audit?”

Sill publishes control mappings to the OWASP LLM Top 10, OWASP Top 10 for Agentic Applications, MITRE ATLAS, and NIST AI RMF. These are mappings, not certifications — Sill holds no security certification today. They can inform your own assessment but do not substitute for an accredited audit.