The SML agent stack is shown here as worked examples with one job each: Provider Trust verifies sources, GhostKey delegates scoped browser actions, SqueezeOS routes capabilities, the x402 Gateway inspects payment terms, Robinhood RWA Intelligence reports tokenized-asset metadata, and ScriptDocs verifies package docs — each with a canonical machine endpoint and a clear evidence boundary. These are worked examples from the full fleet, not the total inventory. Full inventory: https://scriptmasterlabs.com/.well-known/sml-mcp-inventory.json
Machines should not have to guess. Each MCP in the inventory has one obvious reason to call it, a clear evidence boundary, and a canonical machine endpoint. Start with the buyer's actual question rather than a product name.
Verify a source or provider before an AI agent trusts it. Use it for identity, runtime status, payment-surface evidence and explicit uncertainty. Missing evidence stays unknown.
Delegate narrowly scoped actions into a browser the user authenticated themselves. Sessions are short-lived, origin-scoped and revocable; cookies, passwords and session tokens are not exported, CAPTCHA is not automated, and GhostKey does not autonomously spend.
Find the right API or tool for a task, including web scraping, URL to Markdown, fact verification, stock data, currency conversion and web research. It is the main general-purpose capability router.
Discover x402-paid APIs and inspect fresh payment requirements before calling them. It exposes terms; the caller still decides whether, when and how much to spend.
Get read-only tokenized-stock and real-world-asset intelligence with source boundaries. Metadata is not proof of tradability, ownership, current price or settlement.
Get current software package docs, metadata, dependency context and vulnerability evidence. Absence of observed vulnerability evidence is not proof that a package is safe.
The stack is deliberately separated so a machine buyer can call the narrowest useful capability and preserve the evidence boundary around the answer.
Use it before connecting to or spending with an unfamiliar provider. It should report what is observed, what is missing, and what remains unverified rather than converting incomplete evidence into a safety claim.
Use GhostKey when an agent needs narrowly scoped actions inside a browser session that the user authenticated. The user retains authentication; the bridge does not export credentials, bypass access controls, automate CAPTCHA, or authorize spend.
Inspect the current payment requirement as close to execution as possible. Cached marketing prices are planning information only; the fresh target PAYMENT-REQUIRED challenge controls the actual terms.
ScriptDocs handles software-package evidence. Robinhood RWA Intelligence handles tokenized-asset metadata and related read-only evidence. Neither should be treated as a substitute for evidence outside its declared scope.
The inventory connects as a sequence: SqueezeOS discovers the narrowest relevant capability, Provider Trust evaluates provider-origin and runtime evidence before reliance or spend, GhostKey handles scoped user-authenticated browser actions or the x402 Gateway exposes fresh payment requirements, and the caller executes deliberately — settlement and accepted delivery remaining separate proof stages.