These pages publish conceptual contracts, reference schemas, and implementation guidance. They do not claim that a MiRuntime SDK, package, hosted runtime, or certification exists.
Engineering principles
- Keep models replaceable behind capability-aware adapters.
- Keep lifecycle, policy, approvals, tool execution, and evidence in typed software.
- Make side effects bounded, idempotent, checkpointed, and recoverable.
- Preserve context provenance and memory scope.
- Use UTC timestamps and explicit terminal states.
Implementation pathways
Quickstart
One reversible workflow from request to evidence.
02 · Platform engineerRun Lifecycle
Durable states, invariants, concurrency, and terminal behavior.
03 · AI platformModel Adapters
Provider-neutral capability and error contracts.
04 · SecurityTools and Policy
Typed actions, scoped credentials, policy, and approvals.
05 · DataMemory and Lineage
Attributable context, governed memory, and replay facts.
06 · ReliabilityDurable Execution
Resume, retry, restore, compensate, or terminate.
07 · GovernanceObservability and Evidence
Operational signals and reviewer-facing records.
08 · QATesting
Deterministic controls and stochastic outcomes.
09 · ResearchTeleodynamic Profile
Optional research-only goal, viability, reflection, and edit fields.
10 · OperationsDeployment
Atomic releases, rollback, runbooks, and SLOs.
Build sequence
- Define intent and success criteria.
- Define authority and tool contracts.
- Create durable lifecycle state.
- Add model adapters.
- Assemble scoped context.
- Validate proposals.
- Execute bounded side effects.
- Capture evidence.
- Recover or terminate.
- Evaluate the completed workflow.
Role-based reading paths
Application developer
Quickstart → Runtime Contract → Tool Contracts → Testing.
Platform engineer
Lifecycle → Model Adapters → Durable Execution → Deployment.
Security/governance engineer
Tool Policy → Evidence → Implementation Checklist → Teleodynamic Profile boundaries.
Research engineer
Teleodynamic Profile → Benchmarks → Reproducibility → Research program.
Runtime contract
{
"contractVersion": "1.0",
"intent": {
"objective": "Prepare a dependency update",
"riskClass": "reversible-change",
"successCriteria": ["tests-pass", "reviewable-diff"]
},
"context": {
"sources": [],
"memoryScopes": [],
"prohibitedDataClasses": []
},
"authority": {
"allowedTools": [],
"writePolicy": "draft-only",
"requiredApprovals": ["merge", "deploy"]
},
"execution": {
"state": "running",
"checkpointPolicy": "before-side-effect",
"budgets": { "steps": 24, "elapsedSeconds": 900 }
},
"evidence": {
"events": [],
"artifacts": [],
"unresolvedUncertainty": []
}
}Tool registry and policy
Tool contracts define schema, permissions, side effects, credentials, idempotency, timeout, rollback or compensation, approval, and evidence. Policy outcomes are allow, deny, require approval, transform, or escalate.
Testing stochastic workflows
Test deterministic boundaries exactly and probabilistic behavior through repeated trials, allowed outcome ranges, versioned fixtures, fault injection, policy assertions, and evidence completeness.
Reference downloads
Conceptual runtime contract
JSON Schema for intent, context, authority, execution state, budgets, recovery, and evidence.
Download JSON SchemaConceptual tool contract
JSON Schema for typed inputs and outputs, permissions, side effects, credentials, idempotency, retries, and evidence.
Download JSON SchemaConceptual policy decision
JSON Schema for allow, deny, approval, transformation, and escalation decisions evaluated during execution.
Download JSON SchemaConceptual evidence event
JSON Schema for ordered runtime events, actors, status, tools, approvals, artifacts, errors, uncertainty, and integrity metadata.
Download JSON SchemaRun lifecycle schema
JSON Schema for durable run state, sequence, concurrency, checkpoints, and terminal reasons.
Download JSON SchemaModel adapter capability schema
JSON Schema for provider-neutral route identity, model capabilities, placement, and retry behavior.
Download JSON SchemaContext package schema
JSON Schema for attributable sources, hashes, permissions, classifications, freshness, and selection reasons.
Download JSON SchemaCheckpoint schema
JSON Schema for content-addressed checkpoints created before material side effects.
Download JSON SchemaEvaluation case schema
JSON Schema for benchmark versions, model routes, runtime profiles, repetitions, metrics, and limitations.
Download JSON SchemaTeleodynamic runtime profile schema
Research-only JSON Schema for bounded goals, viability metrics, reflection, edits, and immutable constraints.
Download JSON exampleRun lifecycle example
Conceptual JSON example for a reversible dependency-update run.
Download JSON exampleEvidence timeline example
Conceptual ordered event timeline from run creation through completion.
Download JSON exampleTeleodynamic profile example
Research-only JSON example with bounded goals, viability metrics, reflection, and a no-op proposal.
Download JSONReference artifact index
Machine-readable inventory of the reference files distributed with this release.
DownloadThese files are conceptual MiRuntime editorial references, not a product SDK or certification profile.
Source record
References
- Model Context Protocol Specification — 2025-11-25 Primary source
Model Context Protocol project. Model Context Protocol. Published 2025-11-25; last reviewed 2026-06-24 UTC. Protocol specification.
OpenTelemetry project. Cloud Native Computing Foundation. Published Current specification repository; last reviewed 2026-06-24 UTC. Official specification.
National Institute of Standards and Technology. NIST. Published 2023-01-26; last reviewed 2026-06-20 UTC. Government framework.
- Agentic AI Threats and Mitigations Primary source
OWASP Agentic Security Initiative. OWASP. Published Current guidance; last reviewed 2026-06-20 UTC. Security guidance.
