{a}OS Agentic Reference Stack
A vendor-neutral reference model for agentic AI systems — like the OSI model, but for the AI era.
Click a stratum to drill
down
Add products to compare
Filter by vendor or
deployment
Press
? for shortcuts
Map & Rank Your Stack
0Products
7Strata
0Substrates
0Constructs
2Axes
Hover a section
GOV
OBS
L7
Experience & Intent
What did the human actually want — and did they get it?
Substrates
Intent Parsers & UX
Chat interfaces, voice agents, intent classifiers
AF
PC
CI
aE
Constructs
🎯 intent_object
🗓 session_context
✅ satisfaction_signal
🔄 feedback_loop
Generative UI
Dynamic interfaces assembled by agents
V0
LV
Human-in-the-Loop Controls
Approval gates, escalation paths, override mechanisms
GS
L6
Governance & Trust
Who approved this action and under which policy?
Substrates
Policy Engines
OPA, Azure Policy, guardrails-as-code
OP
GS
Constructs
🛡 policy_rule
⚖ policy_evaluation
🚧 guardrail_config
🔄 override_protocol
Identity & Access
Auth, RBAC, credential vaults
Constructs
🔑 auth_token
👥 role_binding
🔐 credential_vault
🎯 permission_scope
Audit & Compliance
Immutable logs, regulatory frameworks
SY
Constructs
📝 audit_entry
✅ compliance_check
📅 retention_policy
🔗 evidence_chain
Safety & Content Filters
Content moderation, toxicity detection, PII redaction
Constructs
🚫 content_filter
🧪 toxicity_scorer
🕶 pii_redactor
🛑 safety_boundary
L5
Observability & Evaluation
How well did this work, and can we prove it?
Substrates
Tracing & Logging
Distributed tracing, structured logs, token accounting
SY
Constructs
🔍 trace_span
📝 log_entry
💰 token_ledger
⏱ latency_histogram
Eval Frameworks
Benchmarks, A/B testing, LLM-as-judge
SP
aE
Constructs
🧪 eval_suite
📋 test_case
⚖ judge_prompt
📊 score_aggregator
Drift & Anomaly Detection
Model drift monitors, SLO alerting, cost anomalies
Constructs
📈 drift_detector
🔔 anomaly_alert
🎯 slo_monitor
💵 cost_tracker
L4
Orchestration & Decisioning
What happens next, in what order, with which agent?
Substrates
Workflow Engines
DAG runners, state machines, orchestration graphs
AF
HM
GD
BM
Constructs
💠 execution_plan
🔄 state_checkpoint
🔗 dependency_graph
⏱ timeout_policy
📦 context_pack
Primitives
← from GSD
active_task_id
repo_slice
included_file_list
excluded_context_list
current_goal
current_constraints
dependency_list
next_action
context_budget_pct
reset_threshold_flag
carry_forward_note
working_assumption
📁 state_file_set
Primitives
← from GSD
state_file_path
checkpoint_id
artifact_pointer
role_tag
progress_flag
resume_token
last_successful_step
last_failed_step
branch_context
handoff_note
📝 product_requirement_doc
Primitives
← from BMAD
requirement_id
feature_statement
user_problem
business_goal
acceptance_criteria_line
scope_flag
non_goal
dependency
assumption
open_question
risk_note
priority_rank
📊 requirement_matrix
Primitives
← from BMAD
requirement_row_id
feature_id
owner
priority
dependency_link
testability_flag
implementation_status
change_risk
clarification_status
approval_status
Agent Frameworks
Multi-agent coordination, task delegation, crew patterns
CW
LG
BM
Constructs
🤖 agent_roster
🎯 task_assignment
💬 message_channel
🛑 delegation_policy
🤝 persona_handoff_packet
Primitives
← from BMAD
persona_name
persona_role
handoff_from
handoff_to
task_objective
constraints_block
unresolved_question
recommended_next_step
decision_owner
escalation_flag
Routing & Planning
LLM routers, cost-aware selectors, plan generators
Scheduling & Triggers
Cron jobs, event listeners, webhook handlers
Retry & Recovery
Circuit breakers, fallback chains, checkpoint/resume
GD
Constructs
✅ lifecycle_checkpoint
Primitives
← from GSD
checkpoint_name
trigger_threshold_pct
purge_flag
reload_file_list
completion_gate
rollback_point
validation_status
operator_override_flag
confidence_note
retry_policy
Quality Gates & Verification
Test gates, eval checkpoints, TDD enforcement, approval
barriers
SP
Constructs
🛑 test_gate
Primitives
← from Superpowers
test_case_id
test_name
pass_fail_bool
coverage_pct
block_forward_flag
failure_trace_id
required_fix_note
gate_status
retry_count
last_run_timestamp
🔄 tdd_workflow_path
Primitives
← from Superpowers
spec_seed
test_generated_flag
implementation_started_flag
failing_test_count
passing_test_count
checkpoint_status
proceed_permission
rollback_flag
review_required_flag
final_gate_status
L3
Execution & Interfaces
How does the agent act on the world?
Substrates
Tool Registries & MCP
Tool catalogs, Model Context Protocol, API gateways
EX
Constructs
🔧 tool_manifest
🔗 api_binding
🛠 capability_schema
Code Execution
Sandboxes, interpreters, REPL environments
SP
CI
Constructs
💻 sandbox_runtime
💬 interpreter_session
📋 execution_result
🚧 resource_limit
🖥 terminal_session
Primitives
← from Superpowers
session_id
pane_id
command_string
working_directory
process_status
exit_code
stdout_ref
stderr_ref
branch_name
checkpoint_marker
External APIs
REST/GraphQL connectors, webhook emitters, data sinks
HA
Constructs
🔗 api_connector
📡 webhook_emitter
⏱ rate_limiter
💾 response_cache
Browser & UI Automation
Playwright, Puppeteer, screen scraping, RPA bridges
Constructs
🌐 browser_session
🖱 page_action
🎯 element_selector
📷 screenshot_capture
File & OS Interfaces
Filesystem access, shell execution, system commands
GI
SY
Constructs
📄 file_handle
💻 shell_command
📂 directory_walker
🔒 permission_check
Message Queues & Events
Pub/sub, event buses, async message passing
Constructs
📬 message_topic
🔔 event_subscriber
📪 dead_letter_queue
✅ delivery_guarantee
L2
Knowledge & Memory
What does the system know, and how does it recall?
Substrates
RAG Pipelines & Search
Vector search, hybrid retrieval, re-ranking
AS
EX
GI
Constructs
🔍 search_index
📄 chunk_collection
🎯 relevance_score
🔄 rerank_pipeline
Memory Systems
Short-term context, long-term persistence, episodic recall
M0
HA
Constructs
🧠 working_memory
📚 episodic_store
🔄 memory_consolidator
🗑 forgetting_policy
Knowledge Graphs
Entity relationships, ontologies, semantic networks
GD
BM
Constructs
📍 entity_node
🔗 relationship_edge
📜 ontology_schema
🔍 graph_query
Embedding Stores
Vector databases, index management, chunking strategies
Constructs
📦 vector_index
🧬 embedding_model
✂ chunk_strategy
🔄 index_refresh
L1
Models & Infrastructure
What is the raw intelligence, and where does it run?
Substrates
Foundation Models
GPT-4o, Claude, Gemini, Llama, Mistral, fine-tunes
AF
HA
Constructs
📇 model_card
🌐 model_endpoint
📏 context_window
🔤 tokenizer_config
Compute & Serving
GPU clusters, serverless inference, model endpoints
Constructs
☁ inference_server
💻 gpu_pool
📦 batch_queue
📈 autoscaler_policy
Training & Fine-tuning
RLHF pipelines, LoRA adapters, distillation
Constructs
🏃 training_run
📊 dataset_config
🧬 lora_adapter
🔄 rlhf_pipeline