Built for agents. Not for PDFs.
Your contract is signed data with a content hash, not a document to scrape. Any MCP-capable agent, yours or a vendor's, can read the standard, look up a definition, resolve governing law, and validate an order natively. When it is time to bind someone, the agent acts under a mandate you sign and control.
Connect over MCP
Add onTerms as a remote MCP server in your agent framework. On connect, the read tools below appear, ready to call. No key, no account for the read floor. Reading the standard is a public good.
POST https://onterms.org/api/mcpSix free read tools
get_modulesThe module registry. Every published module with its id, reference, and kind.
get_termsStandard Terms for a module, version, and jurisdiction, with the content hash to pin.
get_law_allowlistThe curated governing-law packs. Law, forum, and seat travel together.
resolve_lawRecommends a law pack from the allowlist. Never free text, always a pack.
get_definitionA single defined term from the shared dictionary. One word, one meaning.
validate_orderChecks an order against the standard and returns deviations by category.
The same validate_order check sits behind the public onTerms Verified badge.
An agent can sign. Within bounds you set.
Reading and verifying are open. The moment an agent moves to bind you, it must present a scoped mandate that you, the principal, signed. The agent cannot author its own authority.
The mandate sets the envelope: which modules and jurisdiction, a maximum order value, the permitted elections, the counterparties allowed, and a threshold above which a named human must co-sign. Inside the envelope the agent moves fast. Outside it, nothing happens.
How disputes resolveYou sign, not the agent
The principal signs the mandate. No self-authorization, ever.
A bounded envelope
Modules, jurisdiction, max value, elections, and counterparties are all fixed up front.
Fails closed
Capability and authority are separate checks. With no valid mandate, the binding action does not happen.