Orchestrator API
The HTTP API your backend uses to run the lifecycle: register policies, submit intents, submit proposals, and subscribe to outcome events.
delta mandate
Intent enforcement for delegated payments.
delta mandate is a cryptographic enforcement system for financial actions delegated to AI agents. It is built for an operator — a platform that runs agents for its users — to ensure each proposed action respects the user’s intent. Typical operators include agentic commerce platforms and neo-banks.
The user defines an intent. An agent proposes a candidate solution. delta mandate extracts evidence from the proposal, evaluates it against the intent, and produces a cryptographic proof of the evaluation result that anyone can independently verify.
If you are evaluating delta mandate, start with the mandate verification gap. If you are integrating it, start with the integration overview.
How it works
Section titled “How it works”In practice, delta mandate sits between the agent’s proposed action and the operator’s execution layer: your system submits the proposal, delta returns a success/failure result, and the action only proceeds if it satisfies the user’s authorized policy.
- Define — the user decides the intent: the intended outcome plus any required values.
- Author — an agent drafts a policy encoding the intent via the MCP; the user authors an intent referencing that policy, and the operator submits it to the orchestrator.
- Propose — the agent submits a proposal, e.g. a specific product variant.
- Extract — the orchestrator extracts the evidence (relevant facts) from the proposal.
- Verify — the orchestrator evaluates the evidence against the intent and generates a ZK proof.
- Outcome — a
success/failureverdict comes back by stream or poll. Afailurealways says why, and names the failed constraints when constraints caused it.
The managed model
Section titled “The managed model”delta runs the servers. You don’t deploy or operate the orchestrator, the prover, or the internal services. You integrate against exactly two surfaces:
- the Orchestrator API (HTTP / OpenAPI), and
- the policy-generation MCP (for agents).
Signing is optional. An intent carries an authorship mode: signed, or unsigned. The orchestrator does not require a signature, so where you send unsigned intents authorship rests on your own authentication of the user. The mode changes only who is provably the author — evidence extraction, policy evaluation, and the proof over that evaluation are identical either way. See the Orchestrator API.
The two surfaces
Section titled “The two surfaces”MCP
The agent-facing surface for authoring the policy a user’s intent is built on: a language reference, taxonomy search, and policy validation.
Evidence extraction
Section titled “Evidence extraction”Extracting evidence is what turns a proposal into the facts the orchestrator checks. The verification — and the proof it produces — is equally rigorous in every case; what changes is how hard the extraction is, depending on where the data lives and how well-structured it is.
- Available today Sources on UCP — a standard for exposing structured product data. We run a production-ready extraction engine for these.
- Built with you Your own systems — when the facts live in your systems, we build the extraction integration together. Get in touch.
- On the roadmap Unstructured web content — free-form data anywhere on the open internet.
Live previews
Section titled “Live previews”Son of Anton is a hosted live preview, named after the rogue AI in Silicon Valley that buys what no one asked for. An agent shops on a user’s behalf under a guardrail, and delta mandate verifies each proposed product against the real Shopify UCP catalog before the purchase can proceed. Nothing to install — it needs only a browser.
Bring your own agent is the self-serve version for Claude Code, connect to our public MCP and run the full loop yourself: author a policy, propose a product, and read back the verdict.
Son of AntonBring your own agent
Where to go next
Section titled “Where to go next”Evaluating delta mandate
What it enforces, what a mandate can express, and what delta covers when its own engine is wrong.
Integrating it
The two surfaces your backend and your agent call, the policy language, and the generated endpoint reference.