Multi-Agent Lifecycle Protocol (MPLP)

The Agent OS Protocol

The Lifecycle Protocol for AI Agents

Observable. Governed. Vendor-neutral.

Frameworks help you build agents. Protocols ensure agents work together safely and consistently.

If HTTP is how documents travel across the internet, MPLP is how work travels between agents.

Documentation • Specifications • Golden Flows • Schemas • SDKs

---

1. What Is an "Agent OS"?

Modern LLM agents behave like operating system processes — they plan, act, update state, collaborate, and execute long-lived work.

A real Agent OS must define:

Lifecycle • Governance • State • Observability

Frameworks ≠ OS. Runtimes ≠ OS. Only a protocol can define the OS layer.

MPLP is that protocol.

---

2. Why an Agent OS Must Be a Protocol

Operating systems are contracts, not implementations:

• POSIX
• TCP/IP
• SQL
• Kubernetes API

Frameworks implement behavior. Protocols define invariants.

Only a protocol can ensure:

• Vendor neutrality
• Semantic lifecycle guarantees
• Reproducible reasoning
• Cross-framework interoperability
• Stable execution semantics
• Portable agent state

This is why the Agent OS layer must be a protocol, not a tool.

---

3. Why MPLP Exists

Today's agent ecosystem fails for structural reasons:

• No lifecycle semantics
• No governance
• No state model
• No observability
• No reproducibility
• No multi-agent correctness
• No cross-framework semantics

Frameworks give you execution convenience. MPLP provides lifecycle correctness.

---

4. MPLP Architecture (L1 → L4)

---

5. L1 — Core Protocol

Defines:

• lifecycle envelopes
• execution semantics
• reasoning stages
• governance hooks
• trace invariants
• semantic identity

This is the OS contract every runtime must follow.

---

6. L2 — Coordination & Governance Modules (10 Modules)

Module	Description
Context	Initialize lifecycle constraints & objectives
Plan	Deterministic reasoning & orchestration intent
Confirm	Governance, permissions, risk scoring
Trace	Replayable reasoning & action audit
Role	Persona & capability definitions
Dialog	Structured reasoning boundaries
Collab	Multi-agent workflow semantics
Extension	Safe extensibility model
Core	Identity, invariants, protocol constants
Network	External IO under protocol semantics

---

7. L3 — Runtime Glue (AEL / VSL / PSG)

AEL — Action Execution Layer

This is the OS execution loop.

---

VSL — Value State Layer

The semantic state substrate for:

• scoring
• permissions
• governance
• lifecycle valuation
• capability models

---

PSG — Project Semantic Graph

The OS-level semantic filesystem for:

• intents
• plans
• deltas
• documents
• code
• traces

Prevents drift. Ensures reproducibility.

---

8. Cross-Cutting Kernel Duties (OS-Level Duties)

These 11 kernel obligations apply across every lifecycle stage, agent, and runtime. They ensure multi-agent systems remain coherent, auditable, recoverable, and deterministic — the core requirements of an Agent OS.

---

---

9. Execution Profiles

Profile	Purpose
SA	Deterministic single agent
MAP	Governed multi-agent collaboration

---

10. Positioning in the Ecosystem

flowchart TD
    A[MPLP<br/>Protocol Layer - Agent OS]:::protocol
    B[MCP<br/>Communication Layer]:::protocol
    C[Frameworks<br/>LangGraph / AutoGen / CrewAI / Swarm]:::framework
    D[AgentOS Products<br/>PwC / Agno / AG2 / Builder.io]:::runtime

classDef protocol fill:#1f77b4,color:#fff;
classDef framework fill:#2ca02c,color:#fff;
classDef runtime fill:#d62728,color:#fff;

Loading

Framework = App Runtime = Engine MPLP = OS Protocol

---

11. Schemas

Located in /schemas:

mplp-context.schema.json
mplp-plan.schema.json
mplp-confirm.schema.json
mplp-trace.schema.json
mplp-role.schema.json
mplp-dialog.schema.json
mplp-collab.schema.json
mplp-extension.schema.json
mplp-core.schema.json
mplp-network.schema.json

---

12. Compliance

A runtime is MPLP-compliant only if it implements:

• AEL / VSL / PSG
• All 10 modules
• All 11 cross-cutting concerns
• SA & MAP profiles
• Governance shells
• Drift detection
• Replayable trace

---

13. SDKs

TypeScript

@mplp/core
@mplp/schema
@mplp/modules
@mplp/coordination
@mplp/compliance
@mplp/devtools
@mplp/runtime-minimal
@mplp/sdk-ts

Python

mplp-core
mplp-schema
mplp-modules
mplp-runtime

Examples include SA, MAP, drift detection, delta-intent, governance flows.

---

14. Documentation

https://coregentis.github.io/MPLP-Protocol/

---

15. Status & Governance

Version: v1.0.0 (Frozen Spec)
Governance: MPLP Protocol Governance Committee (MPGC)
License: Apache-2.0

Any normative breaking change requires a new protocol version.

---

16. Contributing

Contributions are welcome.
Please see CONTRIBUTING.md for submission process and coding standards.

---

17. License & Copyright

This project is licensed under the Apache License, Version 2.0.
You may not use this project except in compliance with the License.
You may obtain a copy of the License at:

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.

Copyright © 2025 Bangshi Beijing Network Technology Limited Company.
Licensed under the Apache License, Version 2.0.