Protocol Route Planner prototype grant

Summary

The Protocol Route Planner takes an agent's intent and surrounding context and recommends one or more viable routes across agentic-commerce protocols. It is a planning model. It does not execute transactions, take custody of funds, or stand in as a settlement layer. The output is a structured recommendation with explicit conditions, credential requirements, settlement options, and audit trail expectations that downstream systems can consume.

Why It Matters

Agents face a fragmented commerce landscape: ACP, UCP, AP2, x402, and merchant-native flows each carry their own credential and settlement assumptions. Without a shared planning model, every orchestrator re-implements the same decision logic, often inconsistently. A reference planner gives developers a clean separation between "what route should we use" and "how do we execute that route" and lets policy live in one place.

Inputs and Outputs

Strategic Layer Mapping

This grant advances Layer 2: Protocol Routing in the Foundation's seven-layer agentic-commerce development program. It consumes Layer 1 (identity) and Layer 3 (merchant readiness) inputs and feeds Layer 5 (mission compliance reporting).

Strategic Gap Mapping

Closes Gap 2: Cross-Protocol Routing. Merchants and agents speak different commerce protocols. A shared planning model reduces integration cost on both sides and makes policy enforcement portable.

Suggested Deliverables

• Open-source planner library with a documented decision interface.
• Schema definitions for inputs and outputs.
• Reference rule sets covering at least two protocols, for example ACP and x402, plus a documented extension pattern.
• Reference fixtures for happy path, ambiguous path, and unsupported path scenarios.
• Documentation explaining why the planner is planning-only and how it composes with separate execution and settlement systems.
• An optional visualization that renders a planning decision and its branches.

MVP Expectations

• Given a structured input, the planner returns a deterministic structured output.
• At least two protocol rule sets are represented.
• Unsupported conditions are explicitly returned with reason codes, not silently dropped.
• The planner does not execute transactions, hold credentials, or take custody.
• Passing test suite and CLI invocation example.
• Published under an OSI-approved license with a permissive default.

Out of Scope for Initial Grant

• Transaction execution, settlement, or fund custody.
• Wallet, key, or credential storage.
• Hosted SaaS planner endpoints.
• Closed protocol-specific tooling without an extension point for others.
• Regulatory routing classifications or jurisdiction-specific legal advice.

Security and Privacy Expectations

• Planner outputs are deterministic given the same input, with versioned rule sets.
• No sensitive credential material is logged at default verbosity.
• Inputs are validated against the schema.
• A documented threat model that lists what the planner does and does not protect against.
• A responsible disclosure contact in the README.

Suggested Applicant Profile

This grant suits a team with experience in policy engines, payments routing, or protocol-agnostic SDK design. Familiarity with agentic-commerce protocols such as ACP, UCP, AP2, or x402 is helpful. Prior open-source delivery is a plus. The Foundation welcomes both engineering-led teams and research-led teams with a clear delivery partner.

Review Criteria

• Public-good output and permissive licensing.
• Determinism and clarity of the planning interface.
• Quality of the included rule sets.
• Extensibility for additional protocols.
• Quality of documentation and developer ergonomics.
• Maintenance plan and team track record.

How to Apply