The trust layer for AI agents that spend real money.
AI agents are starting to spend on people's behalf. Zado is the consumer-side policy and consent layer that decides what those agents are allowed to do — before any settlement rail moves money.
"We review the budget on the weekends and our agents shop all week. This is not today's reality but it will be the reality in the next 3 years."
What Zado actually is
For people
An ADHD-friendly money app where your agents shop all week and you review on the weekend. Envelope budgeting, an AI coach, and a human approval gate that asks before anything important moves.
For agents
An open, MCP-native consent surface. Any agent that speaks MCP can ask Zado "can I spend $87 on groceries for my user" and get a policy answer back, with a human-approval loop for anything above threshold.
For partners
An AP2 v0.2 Trusted Surface in the Trusted Agent Provider model — the standardized role for the consumer-side trust layer in agent commerce. AP2 was donated to the FIDO Alliance in October 2025.
The Human Approval Gate
The keystone product surface. When an agent wants to spend more than the user wants to delegate without asking, Zado pauses the transaction and asks:
Your agent wants to spend $250.00 on groceries and your budget says you have $200.00 allowed for the week. Do you want to allow this — and if so do you want me to set up a reminder to move money in the budget?
Each clause is doing real work. It names the agent. It frames envelope state in the user's actual time unit. It asks consent without shame. And it offers to handle the consequence — which no card network will ever build, because it requires the coach voice and budget context that lives in Zado.
An open protocol, not a closed app
The Zado Agent Trust Protocol is published openly under CC-BY 4.0 because the consumer-side policy layer for agent commerce needs to be open or it won't get adopted. The reference implementation (this app) is where the value lives; the spec itself is for the broader ecosystem to use.
- Protocol v0.2 (current draft) — two-phase delegated spend with the authorize-then-claim pattern, AP2 v0.2 Trusted Surface mapping, and the 6-tool MCP surface
- Protocol v0.1 — the initial published spec
Agent integration is one line: pip install zado-mcp
Built by Evolving Intelligence AI
Zado is one of several checkpoint products from Evolving Intelligence AI — a small team building infrastructure for the agentic-AI era. Other work in progress: Sentinel (project intelligence for AI-assisted development), Plumbline (write-time verification for AI-generated code), Morpheus (phase-gate enforcement for AI dev loops).
If you're a partner in the agent-commerce, agent-trust, or consumer-finance-infrastructure space and want to talk: nicholas.smith@evolvingintelligence.ai.