State Drift
Agents lose intent and constraints as state transitions occur without formal invariants.
Protocol Specification v1.0.0
Multi-Agent Lifecycle Protocol
The Agent OS Protocol
The Lifecycle Standard for AI Systems
MPLP is a vendor-neutral lifecycle protocol for AI agent systems. It makes plans, confirmations, and traces observable and comparable across implementations—without tying you to any framework or vendor.
Not a framework. Not a runtime. Not a platform.
Status
Frozen / Stable
License
Apache 2.0
Org
MPGC Governed
Protocol Status: Frozenv1.0.0-stable
The Governance Gap
The bottleneck is not intelligence. It is lifecycle governance.
Multi-agent systems fail not because agents are weak, but because lifecycle semantics are undefined. These failures are structural outcomes of missing protocol-level invariants.
Frameworks scale features.
Protocols scale ecosystems.
Hallucination Accumulation
Errors compound across agent boundaries due to missing semantic validation frames.
Orchestration Collapse
Coordination complexity grows exponentially without a unified lifecycle protocol.
Audit Black Holes
Traceability is lost when lifecycle events are not governed by a canonical standard.
Protocol Topology
A Protocol Stack, Not a Framework Stack
MPLP sits above agent frameworks and below applications, defining lifecycle semantics that conformant systems are expected to respect. Formal definitions live in the documentation.
Core Protocol
Lifecycle primitives and semantic invariants.
Coordination
Governance primitives: Context, Plan, Confirm, Trace.
Execution
AEL loops, VSL logic, and Project Semantic Graph.
Integration
Models, tools, and external system adapters.
Protocol Modules
Modules are not features. They are lifecycle constraints.
Each module defines lifecycle constraints and evidence surfaces. Systems may adopt modules incrementally, but MPLP remains a protocol specification—not a framework.
Trace (Evidence Surface)
Review the event model and replayable audit surface.
Confirm (Governance Gate)
Understand confirmation gates and permission boundaries.
Plan & Context (Lifecycle Semantics)
Learn how intent, plans, and constraints are represented.
Orientation
Understand MPLP in 5 minutes — then evaluate with evidence.
MPLP defines lifecycle invariants and evidence formats (Plan/Confirm/Trace/Snapshots) so evaluations are comparable across substrates.
This website provides discovery and positioning only. Normative specifications and schemas live in the documentation; the repository is the source of truth.
Core Protocol Flows
Protocol Conformance
Evidence-Based Verification
Unlike static frameworks, MPLP conformance is verified through dynamic Golden Flows—predefined scenario execution that produces replayable evidence of protocol lifecycle adherence.
System providers execute these flows within a protocol-conformant runtime to generate Proof of Governance (PoG) reports.
Learn Conformance ModelProfile Suites
SA Profile (2)
MAP Profile (2)
Suites are scenario groupings for evidence-based self-evaluation, not certification categories.
Ecosystem Topology
Built for Architects, Developers, and Auditors
The specification describes a structural approach for building observable and auditable agent systems. Canonical SDKs and schemas ensure rapid, safe integration.
Governance
Evidence Chain
Plan → Confirm → Trace: audit-ready evidence for agent systems through replayable lifecycle artifacts.
Explore Governance
Resources
SDKs & Schemas
Canonical SDKs and module schemas to implement MPLP incrementally without vendor lock-in.
Explore Ecosystem
Evolution
Versioning Policy
Transparent evolution policy for the frozen v1.0.0 specification and forward-compatible extensions.
View Governance
Evidence
Validation Lab
Browse recheckable, evidence-based adjudications. Coverage reflects evidence maturity — not endorsement, ranking, or certification.
Live counts and verdict details are shown in the Lab.
Open Validation Lab
Standards Alignment (Informative)
Enterprise Evaluation Signals
Reference mappings to external governance frameworks. MPLP lifecycle evidence artifacts may support documentation and auditability practices.
AI Management
ISO/IEC 42001
Reference mapping to lifecycle governance objectives and auditability expectations.
View Alignment
Risk ManagementNIST AI RMF
Reference mapping to GOVERN / MAP / MEASURE / MANAGE functions.
View Alignment
High-Risk AIEU AI Act
Reference mapping to Articles 9–15 documentation-oriented requirements.
View Alignment
Informative mapping only. No certification, endorsement, or legal advice. All regulatory determinations remain the responsibility of the adopting organization.
Build against the Specification.
Build agent systems that remain reliable, observable, and governable — even as models, frameworks, and vendors change.
Protocol v1.0.0•Apache 2.0•MPGC Governed
Positioning Notice: This website provides discovery and positioning content only. For formal protocol definitions, see docs.mplp.io. Source of truth: GitHub Repository.