Module backend 

Unified execution backend for AI agent tasks.

Provides the [AgentBackend] trait that abstracts over different execution modes (direct API, CLI subprocess, simulated) behind a single interface. Unified execution backend for AI agent tasks.

Provides the [AgentBackend] trait that abstracts over different execution modes:

• Direct API calls to LLM providers
• CLI subprocess spawning (Claude Code, OpenCode, Cursor)
• Simulated execution for testing and dry-runs

All callers interact through the same interface regardless of how the underlying LLM execution happens.

Modules

cli

CLI subprocess backend.

direct

Direct API backend.

simulated

Simulated backend for testing and dry-runs.

Structs

AgentHandle

Handle to a running agent execution.

AgentRequest

Request to execute an agent task.

AgentResult

Result of an agent execution.

TokenUsage

Token usage statistics.

ToolCallRecord

Record of a single tool call during execution.

Enums

AgentEvent

Events emitted during agent execution.

AgentStatus

Status of an agent execution.

Traits

AgentBackend

A backend that can execute an AI agent task.

Functions

create_backend

Create a backend from a harness specification.

create_simulated_backend

Create a simulated backend for testing and dry-runs.