dpp

Project intro

dpp (Decision Provenance Protocol) is an experimental Rust CLI built for a very 2026 problem: how do you verify what an AI agent actually decided, and who authorized it? The answer it proposes is signed, verifiable receipts for every multi-agent decision — a provenance record you can check offline, with no network calls, no external services, and (in its author's words) "no crypto/blockchain stuff." Why it's worth a look: the tool's design and its Rust implementation were built start-to-finish with Claude Code — from brainstorming the protocol in spec.md and plan.md down to the actual implementation — which makes it a genuine stress test of agents on real systems work, not just CRUD apps. The feature list reads like a builder who has felt the pain: offline-first verification, agent-friendly CLI/JSON output, scoped permissions, tamper-evident receipts. Status is experimental (0 stars, MIT licensed), so treat it as a promising proof-of-concept from the Claude community showcase rather than production tooling. But the problem it's attacking — accountability in multi-agent workflows — is going to be one of the defining developer-tools questions of the next few years.