Struct OpenCodeRunner 

pub struct OpenCodeRunner;

Adapter for OpenCode CLI (opencode binary).

Headless invocation: opencode run –format json “”

The run subcommand is non-interactive: all permissions are auto-approved, and the process exits after the task completes.

With --format json, output is NDJSON with event types:

• { type: “step_start”, sessionID, part: { type: “step-start”, … } }
• { type: “text”, sessionID, part: { type: “text”, text, … } }
• { type: “tool_use”, sessionID, part: { type: “tool”, callID, tool, state: { status, input, output, … } } }
• { type: “step_finish”, sessionID, part: { type: “step-finish”, reason, cost, tokens: { input, output, cache: { read, write } } } }

Trait Implementations

impl AgentRunner for OpenCodeRunner

fn name(&self) -> &str

Human-readable name of this agent backend.

fn is_available(&self) -> bool

Check whether the agent binary is available on the system.

fn binary_path(&self, config: &TaskConfig) -> Result<PathBuf>

Resolve the binary path (user override or PATH lookup).

fn build_args(&self, config: &TaskConfig) -> Vec<String>

Build the command-line arguments for the agent process.

fn build_env(&self, _config: &TaskConfig) -> Vec<(String, String)>

Build the environment variables for the agent process.

fn run<'life0, 'life1, 'async_trait>( &'life0 self, config: &'life1 TaskConfig, cancel_token: Option<CancellationToken>, ) -> Pin<Box<dyn Future<Output = Result<StreamHandle>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait,

Run the task and return a StreamHandle with event stream and cancel token.

fn capabilities(&self) -> AgentCapabilities

What features this agent supports.

fn version(&self, config: &TaskConfig) -> Option<String>

Get the version of the installed agent binary.

fn validate_config(&self, config: &TaskConfig) -> Vec<ConfigWarning>

Validate config against this agent’s capabilities.