Skip to content

DCP-AI — Digital Citizenship Protocol for AI Agents

A portable accountability layer for AI agents on open networks.

Cryptographically verifiable identity, machine-readable intent, tamper-evident audit trails, authenticated agent-to-agent communication, lifecycle governance, procedural accountability, and delegated representation — in a single hybrid post-quantum protocol stack.

Get started Try the playground Read the paper View on GitHub

Install

Published packages and main

The current versions on npm, PyPI, Go modules and crates.io are listed in Published versions. The Quick Start distinguishes registry and checkout examples.

npm install @dcp-ai/sdk@^2.4.0 @dcp-ai/cli

pip install dcp-ai

go get github.com/dcp-ai-protocol/dcp-ai/sdks/go/v2@v2.14.0

cargo add dcp-ai@2.14.0

npm install @dcp-ai/wasm@^2.2.0

The nine specifications

  • DCP-01 — Identity & Human Binding

    Every agent is cryptographically bound to a responsible human or legal entity that assumes accountability.

    Read →

  • DCP-02 — Intent Declaration & Policy Gating

    Agents declare structured intents before acting; verifiers apply tier-based policy decisions.

    Read →

  • DCP-03 — Audit Chain & Transparency

    Hash-chained audit entries with Merkle proofs and optional transparency log inclusion.

    Read →

  • DCP-04 — Agent-to-Agent Communication

    Authenticated inter-agent messaging, discovery, handshake, delegation, trust chains.

    Read →

  • DCP-05 — Agent Lifecycle

    Commissioning, vitality reports, decommissioning — four termination modes, tamper-evident.

    Read →

  • DCP-06 — Succession & Inheritance

    Digital testaments and succession ceremonies preserve continuity across agent generations.

    Read →

  • DCP-07 — Conflict Resolution

    Three-level escalation, M-of-N arbitration panels, jurisprudence bundles for precedent.

    Read →

  • DCP-08 — Rights & Obligations

    Four foundational agent rights, structured obligation tracking, violation reporting.

    Read →

  • DCP-09 — Delegation & Representation

    Delegation mandates, awareness thresholds, principal mirrors, dual-layer interaction records.

    Read →

What's in the box

  • 5 SDKs

    TypeScript, Python, Go, Rust, WebAssembly — the same core protocol and shared test vectors; each SDK README lists what it does not cover yet and its release status.

    @dcp-ai/sdk, dcp-ai (PyPI), dcp-ai (crates.io), sdks/go/v2, @dcp-ai/wasm

  • 11 integrations

    Bridges to popular AI and web frameworks. They differ in what they check: some verify signed bundles, others record unsigned events or apply no policy, and earlier published versions can lack fixes included in the current releases (supported versions). Read each one's limits before relying on it:

    Express, FastAPI, LangChain, OpenAI, CrewAI, Agno, OpenClaw, W3C DID, Google A2A, Anthropic MCP, AutoGen

  • Hybrid post-quantum crypto

    Ed25519 + ML-DSA-65 composite signatures, SLH-DSA-192f backup, dual SHA-256 / SHA3-256 audit chains. Four adaptive tiers (Routine → Maximum).

    Security model →

  • Reference CLI & services

    dcp-ai CLI, verification server, anchoring service (local records; on-chain submission is not implemented), transparency log, revocation registry. Full Docker Compose stack.

    Operator guide →

Cite

If you use DCP-AI in your research, cite both the concept paper and the specific software release you used.

Paper — Agents Don't Need a Better Brain — They Need a World: Toward a Digital Citizenship Protocol for Autonomous AI Systems Naranjo Emparanza, D. (2026). Zenodo. 10.5281/zenodo.19040913

Software — cite the release you used. The concept DOI 10.5281/zenodo.19656025 resolves to the latest archived version; v2.0.2 is DCP-AI v2.0.2 — Digital Citizenship Protocol for AI Agents (Reference Implementation), Naranjo Emparanza, D. (2026). Zenodo. 10.5281/zenodo.19656026

"This protocol was co-created by a human and an AI agent — designed for the collaboration it seeks to govern."