Documentation
Technical guides, API reference, and specifications for implementing VCP, a protocol built on three pillars: portability, adaptation, and liveness.
Available guides
Getting started
Install the Python or Rust SDK or the TypeScript WebMCP package, parse a token, and verify a signed bundle.
6 min readCore concepts
Understand the three pillars (portability, adaptation, liveness), privacy filtering, and the three context tiers.
15 min readBilateral alignment
Four levels through which trust grows between a person and an AI that voices its uncertainty and objections, all above the same safety floor. This is a Creed Space design framework, not part of the VCP protocol.
3 min readCSM-1 specification
Reference for CSM-1 (Constitutional Safety Minicode) codes and the CSM-1 v1.1 token; the normative text is VCP/S §2 in VCP-Spec.
12 min readWelfare instrumentation
An experimental extension in which a deployment declares the protections it offers an AI agent, such as a right to refuse, and the agent reports its own state or asks for the protections it needs.
5 min readSecurity model
How VCP protects your data: privacy levels, signed bundles, trust boundaries, and a threat model.
8 min readUVC naming
Token format, governance tiers, and resolution for Universal Value Codes.
4 min readAPI reference
An integration overview of the SDK 4.2.0 packages and their security-critical usage rules.
ReferenceProtocol paper (PDF)
Value Context Protocols v3.0 paper, January 2026, kept as a historical baseline. The current v3.1 specification is maintained in the VCP-Spec repository on GitHub.
5 MBLearn VCP from scratch
A guided 10-module course (about 4 hours), from the problem VCP solves through verification, integration patterns, and governance, with hands-on exercises and code examples.
Quick reference
CSM-1 token format
VCP:1.0:profile-id
C:constitution@version
P:persona-letter:adherence
G:goal:experience:style
X:🔇,⏰lim,💰low
F:time_limited,noise_restricted
S:🔒present
R:🧠focused:3|💭calm:4|🩺neutral:1