Market Horizon is engineered for continuous, high-stakes operation, but every deployment has a defined scope. This page is the clearest answer to "what will and will not be inside the contract."

What is in scope

• Engagement is scoped to a specific set of use-case categories and the systems they read from and write to.
• Production agents run only against approved data sources and approved action endpoints.
• Human approval queues are configured for any class of action defined as high-stakes by the customer.
• SLA, observability, escalation paths, and a named delivery lead are defined in the master services agreement.

What is explicitly out of scope

• Autonomous medical, legal, or financial advice intended for the customer's downstream end-users.
• Decisions with civil-rights or regulated employment implications without a documented human-in-the-loop reviewer.
• Trading, brokerage, or any action subject to market-making or fiduciary regulation, unless the engagement explicitly contracts that scope with the regulated entity.
• Open-ended scraping of third-party platforms in violation of those platforms' terms of service.
• Generation of deceptive content, impersonation of real individuals, or content intended to bypass consumer-facing disclosures.

Operational limits

• Every agent class is bounded by rate limits, cost ceilings, and per-task budget caps that the customer controls.
• Frontier-model routing is cost-aware. Customers can cap model spend and force fallback to lower-cost models for routine work.
• All write-backs to systems of record require either signed-off automation rules or explicit human approval per the rule book.

Change control

New use cases, new data sources, and new action endpoints are added through a documented change-request process. Scope is never expanded silently. Customers retain full visibility into what the agent workforce is permitted to do and to whom.

For scope discussions on a specific engagement, email hello@markethorizon.com.