II.
AgentCoreImpl overview
Reference · liveagent-core-impl:codex.core@1.x
Codex CLI Core (1.x) overview
Inspect the raw attributes, linked wiki pages, and inbound or outbound graph edges for agent-core-impl:codex.core@1.x.
Attributes
displayName
Codex CLI Core (1.x)
agentVersionId
agentVersion:codex:ge-0-119-0
packageRef
loopIteratorPolicy
tool-use-loop
loopIteratorNotes
Tool-use loop driven by the OpenAI Responses API. Each turn the model emits a `response.output_item.added` event sequence; the runtime executes any tool calls via its bundled shell sandbox and posts back `function_call_output` items. The loop advances until a `response.completed` event is observed.
contextManagementStrategy
rolling-summary
compactionTriggerNotes
Codex maintains a rolling summary of older turns; the cutover is decided by the Codex Rust runtime, not by an exposed token threshold.
subagentInvokerPolicy
tool-call-handoff
subagentInvokerNotes
Codex emits a Responses-API `handoff` event (or a structured `function_call` targeting another agent profile) and the runtime swaps the active profile/model for the duration of the sub-task.
resultSynthesisPolicy
model-direct
resultSynthesisNotes
Final response text is streamed via `response.output_text.delta` events and finalized on `response.completed`; the CLI prints the assistant text to stdout when `--output-last-message` or default streaming mode is used.
stopDetectionStrategy
structured-end-event
transportClientLibrary
codex-internal
supportedTransportProtocols
parallelToolCallHandling
native
streamingFidelity
full
thinkingChannelHandling
model-dependent
Outgoing edges
composes1
- agent:codex·AgentProductCodex CLI
emits_message_type5
- protocol-message:openai-response-created·ProtocolMessageOpenAI Responses / response.created
- protocol-message:openai-response-output-item-added·ProtocolMessageOpenAI Responses / response.output_item.added
- protocol-message:openai-response-output-text-delta·ProtocolMessageOpenAI Responses / response.output_text.delta
- protocol-message:openai-response-completed·ProtocolMessageOpenAI Responses / response.completed
- protocol-message:openai-response-failed·ProtocolMessageOpenAI Responses / response.failed
fires_hook2
- hook-surface:codex.on-tool-call·HookSurfaceOnToolCall (Codex)
- hook-surface:codex.on-stop·HookSurfaceOnStop (Codex)
realizes1
- layer:4-agent-core·LayerAgent-Core
speaks1
- model-transport:openai-responses·ModelTransportProtocolOpenAI Responses
supports4
- capability:parallel-tool-calls·CapabilityParallel tool calls
- capability:requires-tool-approval·CapabilityRequires tool approval
- capability:streaming·CapabilityStreaming
- capability:supports-tool-use·CapabilitySupports tool use
supports_effort_level4
- effort-level:minimal·EffortLevelEffort: minimal
- effort-level:low·EffortLevelEffort: low
- effort-level:medium·EffortLevelEffort: medium
- effort-level:high·EffortLevelEffort: high
Incoming edges
composed_of2
- agentVersion:codex:ge-0-0-0-lt-0-119-0·AgentVersionCodex CLI (pre-0.119)
- agentVersion:codex:ge-0-119-0·AgentVersionCodex CLI