OpenCode Harness

Use agent_harness="opencode_sdk" when you want Managed Research to run through the OpenCode SDK harness.

Launch

run = client.research.runs.start(
    "Review the repo and propose the smallest high-impact fix.",
    host_kind="daytona",
    work_mode="directed_effort",
    providers=[{"provider": "openrouter"}],
    runbook="lite",
    agent_harness="opencode_sdk",
    agent_model="anthropic/claude-haiku-4-5-20251001",
)

OpenCode models

Model ID	Launch access
`anthropic/claude-sonnet-4-6`	Beta Access
`anthropic/claude-haiku-4-5-20251001`	Standard
`x-ai/grok-4.1-fast`	Standard
`moonshotai/kimi-k2.6`	Standard

OpenCode models do not list reasoning_effort support in the backend catalog.

Failure notes

If the requested harness/model pair is unsupported, launch preflight and run start fail with a structured denial. See Models and Harnesses.

Checkpoint Branching

⌘I

Getting Started

Managed Research

Stack

Research Factory

Prompt Optimization

Infrastructure

SDK Reference

Research API

Reference

Cookbooks

Launch

OpenCode models

Failure notes

​Launch

​OpenCode models

​Failure notes

Launch

OpenCode models

Failure notes