Mission Layer — the execution primitive powering every agent on Metaprise. Read the changelog →
AURA OrgKernel Mission Layer Runtime Control Plane
CORE EXECUTION PRIMITIVES

Mission Layer

The fundamental execution unit of the Metaprise Agent OS. Every agent action — from a simple tool call to a multi-step enterprise workflow — is declared, constrained, executed, and recorded as a Mission.

MISSIONDEFINITION SCHEMA
MissionDefinition
objective string What the mission must accomplish
tools array Permitted tools for this execution
authority L1–L4 Required authority level from graph
timeout duration Maximum execution time allowed
6
Primitives
8
State Machine
4
Constraint Groups
Immutable Outcomes

Six primitives.
Complete control.

Every Mission flows through six precisely defined primitives. Together they guarantee that agent execution is declared, constrained, traceable, and immutable — from launch to close.

PRIMITIVE 01

MissionDefinition

Four-Field Declaration

The foundational declaration of what a Mission is: objective, tools, authority level, and timeout. Immutable once the Mission launches — no scope creep, no privilege escalation.

OBJECTIVE TOOLS AUTHORITY TIMEOUT
PRIMITIVE 02

MissionState

Strict 8-State Machine

Every Mission transitions through exactly 8 states: CREATED → PLANNING → WAITING_APPROVAL → APPROVED → PENDING_EXECUTION → IN_PROGRESS → EXECUTED → CLOSED. No shortcuts, no hidden states.

8 STATES DETERMINISTIC AUDITED
PRIMITIVE 03

MissionContext

Immutable Input Environment

The input environment provided to the Mission at launch. Once launched, MissionContext is frozen — the agent cannot modify its own input data. This prevents self-justification loops and data contamination.

IMMUTABLE FROZEN AT LAUNCH READ-ONLY
PRIMITIVE 04

MissionBoundary

Four Constraint Groups

Four constraint groups — authority, policy, data, and runtime — snapshot-frozen at launch. The agent operates within these boundaries for the entire execution. No boundary can be widened mid-flight.

AUTHORITY POLICY DATA RUNTIME
PRIMITIVE 05

MissionOutcome

Immutable Result Record

The final result of the Mission — written once when state transitions to CLOSED, then sealed permanently. Cannot be modified, deleted, or overwritten. The definitive record of what the agent accomplished.

WRITE-ONCE SEALED PERMANENT
PRIMITIVE 06

MissionStateData

Audited Mutable State

The only mutable primitive — the agent's working memory during execution. Every write to MissionStateData generates an AuditChain entry. Full transparency into how the agent's understanding evolved.

MUTABLE AUDIT-ON-WRITE CHAIN-LINKED
DECLARATION & INPUT

MissionDefinition
& MissionContext

A Mission begins with a four-field declaration: what it must do, which tools it may use, what authority it needs, and how long it may run. The input environment — MissionContext — is snapshot-frozen at launch. The agent reads its context but can never modify it.

MISSIONDEFINITION
Objective: Natural-language or structured goal — what the Mission must accomplish
Tools: Explicit allowlist of tools the agent may invoke during execution
Authority: The L1–L4 authority level required, checked against the Authority Graph
Timeout: Hard deadline — if the Mission exceeds this, it transitions to CLOSED with a timeout outcome
MISSIONCONTEXT
Immutable Snapshot: All input data is frozen at launch — no late-binding data injection
Read-Only Access: The agent can reference its context but cannot write to it
Anti-Contamination: Prevents self-justification loops where agents modify their own inputs
Audit Trail: Context snapshot is hash-recorded in AuditChain for full reproducibility
STATE MACHINE

MissionState

Every Mission transitions through exactly 8 deterministic states. Each transition is validated by the Control Plane, recorded in the AuditChain, and visible in real-time via Observability. No state can be skipped; no transition can be forced.

8 STATES EXPLAINED
CREATED: Mission declared — definition and context registered
PLANNING: Agent formulates execution plan within boundaries
WAITING_APPROVAL: Plan submitted for human or policy approval
APPROVED: Approval granted — execution authorized
PENDING_EXECUTION: Queued for runtime resources
IN_PROGRESS: Agent actively executing tools and writing state
EXECUTED: All actions complete — awaiting outcome finalization
CLOSED: MissionOutcome sealed — immutable and permanent
CONSTRAINTS & RESULTS

MissionBoundary
& MissionOutcome

MissionBoundary defines the cage the agent operates within — four constraint groups frozen at launch. MissionOutcome is the sealed result — written once, read forever. Together they guarantee that what the agent was allowed to do, and what it actually did, are permanently recorded.

MISSIONBOUNDARY — 4 CONSTRAINT GROUPS
Authority Boundary: Which Authority Graph levels the agent can access (L1–L4)
Policy Boundary: Which Policy Engine rules apply to this Mission's tool calls
Data Boundary: Which data sources and classifications the agent may read or write
Runtime Boundary: CPU, memory, duration, and concurrency limits for execution
MISSIONOUTCOME
Write-Once: Outcome is written when state transitions to CLOSED — then sealed
Immutable Record: Cannot be modified, overwritten, or deleted after sealing
Full Provenance: Links back to the MissionDefinition, Context, Boundary, and AuditChain
Compliance Evidence: Serves as the definitive record for regulatory and audit purposes

Eight states. Zero shortcuts.

Every Mission follows this exact path. Each state transition is validated, recorded in the AuditChain, and visible in real-time through Observability.

01
CREATED
Definition & context registered
02
PLANNING
Agent formulates plan
03
WAITING
Awaiting approval
04
APPROVED
Execution authorized
05
PENDING
Queued for runtime
06
IN PROGRESS
Actively executing
07
EXECUTED
Actions complete
08
CLOSED
Outcome sealed

Every Mission is fully reproducible: The combination of MissionDefinition, MissionContext, MissionBoundary, MissionStateData, and MissionOutcome — all linked through the AuditChain — means any Mission can be replayed, audited, and verified at any time.

Every agent action — declared, constrained, and sealed.

Six primitives. Eight states. Complete governance over every Mission in your organization.