> ## Documentation Index
> Fetch the complete documentation index at: https://agent-compass.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Legacy Harness Parameters

Earlier Harness parameter names remain available as temporary aliases during migration. Use the current names in new configurations. Equivalent old and current values merge; conflicting values report an error.

<table>
  <thead><tr><th>Harness</th><th>Legacy parameter</th><th>Current parameter</th></tr></thead>

  <tbody>
    <tr>
      <td>
        `core_agent`
      </td>

      <td>
        `install_timeout`
      </td>

      <td>
        `install_timeout_seconds`
      </td>
    </tr>

    <tr>
      <td>
        `core_agent`
      </td>

      <td>
        `step_limit`
      </td>

      <td>
        `max_turns`
      </td>
    </tr>

    <tr>
      <td>
        `core_agent`
      </td>

      <td>
        `primary_model_max_cost_usd`
      </td>

      <td>
        `max_cost_usd`
      </td>
    </tr>

    <tr>
      <td>
        `discovery_agent`
      </td>

      <td>
        `command_timeout`
      </td>

      <td>
        `tool_timeout_seconds`
      </td>
    </tr>

    <tr>
      <td>
        `mini_swe_agent`
      </td>

      <td>
        `command_timeout`
      </td>

      <td>
        `tool_timeout_seconds`
      </td>
    </tr>

    <tr>
      <td>
        `naive_search_agent`
      </td>

      <td>
        `mode`
      </td>

      <td>
        `delegation_mode`
      </td>
    </tr>

    <tr>
      <td>
        `naive_search_agent`
      </td>

      <td>
        `sub_agent_max_iterations`
      </td>

      <td>
        `subagent_max_turns`
      </td>
    </tr>

    <tr>
      <td>
        `naive_search_agent`
      </td>

      <td>
        `sub_agent_concurrency`
      </td>

      <td>
        `subagent_max_concurrency`
      </td>
    </tr>

    <tr>
      <td>
        `naive_search_agent`
      </td>

      <td>
        `max_iterations`
      </td>

      <td>
        `max_turns`
      </td>
    </tr>

    <tr>
      <td>
        `naive_search_agent`
      </td>

      <td>
        `max_retry`
      </td>

      <td>
        `max_attempts`
      </td>
    </tr>

    <tr>
      <td>
        `naive_search_agent`
      </td>

      <td>
        `retry_interval`
      </td>

      <td>
        `retry_interval_seconds`
      </td>
    </tr>

    <tr>
      <td>
        `naive_search_agent`
      </td>

      <td>
        `request_timeout`
      </td>

      <td>
        `request_timeout_seconds`
      </td>
    </tr>

    <tr>
      <td>
        `openai_chat`
      </td>

      <td>
        `max_retries`
      </td>

      <td>
        `max_attempts`
      </td>
    </tr>

    <tr>
      <td>
        `openclaw`
      </td>

      <td>
        `openclaw_version`
      </td>

      <td>
        `version`
      </td>
    </tr>

    <tr>
      <td>
        `openclaw`
      </td>

      <td>
        `max_tokens`
      </td>

      <td>
        `max_output_tokens`
      </td>
    </tr>

    <tr>
      <td>
        `openclaw`
      </td>

      <td>
        `context_window`
      </td>

      <td>
        `context_window_tokens`
      </td>
    </tr>

    <tr>
      <td>
        `openevolve`
      </td>

      <td>
        `python`
      </td>

      <td>
        `python_binary`
      </td>
    </tr>

    <tr>
      <td>
        `openevolve`
      </td>

      <td>
        `iterations`
      </td>

      <td>
        `max_turns`
      </td>
    </tr>

    <tr>
      <td>
        `openhands`
      </td>

      <td>
        `openhands_version`
      </td>

      <td>
        `version`
      </td>
    </tr>

    <tr>
      <td>
        `openhands`
      </td>

      <td>
        `max_iterations`
      </td>

      <td>
        `max_turns_per_run`
      </td>
    </tr>

    <tr>
      <td>
        `openhands`
      </td>

      <td>
        `conversation_timeout`
      </td>

      <td>
        `request_timeout_seconds`
      </td>
    </tr>

    <tr>
      <td>
        `openhands`
      </td>

      <td>
        `command_timeout`
      </td>

      <td>
        `tool_timeout_seconds`
      </td>
    </tr>

    <tr>
      <td>
        `openhands`
      </td>

      <td>
        `terminal_no_change_timeout_seconds`
      </td>

      <td>
        `terminal_idle_timeout_seconds`
      </td>
    </tr>

    <tr>
      <td>
        `openhands`
      </td>

      <td>
        `terminal_max_output_size`
      </td>

      <td>
        `max_terminal_output_chars`
      </td>
    </tr>

    <tr>
      <td>
        `openhands`
      </td>

      <td>
        `enable_condenser`
      </td>

      <td>
        `enable_context_compaction`
      </td>
    </tr>

    <tr>
      <td>
        `openhands`
      </td>

      <td>
        `condenser_max_size`
      </td>

      <td>
        `context_compaction_threshold_events`
      </td>
    </tr>

    <tr>
      <td>
        `openhands`
      </td>

      <td>
        `condenser_keep_first`
      </td>

      <td>
        `context_compaction_keep_first_events`
      </td>
    </tr>

    <tr>
      <td>
        `researchharness`
      </td>

      <td>
        `install_timeout`
      </td>

      <td>
        `install_timeout_seconds`
      </td>
    </tr>

    <tr>
      <td>
        `researchharness`
      </td>

      <td>
        `llm_request_timeout_seconds`
      </td>

      <td>
        `request_timeout_seconds`
      </td>
    </tr>

    <tr>
      <td>
        `researchharness`
      </td>

      <td>
        `compact_trigger_tokens`
      </td>

      <td>
        `context_compaction_trigger_tokens`
      </td>
    </tr>

    <tr>
      <td>
        `scicode_tool_use`
      </td>

      <td>
        `mode`
      </td>

      <td>
        `generation_mode`
      </td>
    </tr>

    <tr>
      <td>
        `scicode_tool_use`
      </td>

      <td>
        `tool_names`
      </td>

      <td>
        `tools`
      </td>
    </tr>

    <tr>
      <td>
        `scicode_tool_use`
      </td>

      <td>
        `tool_use_max_loops`
      </td>

      <td>
        `max_turns_per_step`
      </td>
    </tr>

    <tr>
      <td>
        `scicode_tool_use`
      </td>

      <td>
        `code_timeout_seconds`
      </td>

      <td>
        `tool_timeout_seconds`
      </td>
    </tr>

    <tr>
      <td>
        `scicode_tool_use`
      </td>

      <td>
        `sandbox_timeout_seconds`
      </td>

      <td>
        `tool_timeout_seconds`
      </td>
    </tr>

    <tr>
      <td>
        `scicode_tool_use`
      </td>

      <td>
        `sandbox_memory_limit_mb`
      </td>

      <td>
        `sandbox_max_memory_mib`
      </td>
    </tr>

    <tr>
      <td>
        `scicode_tool_use`
      </td>

      <td>
        `sandbox_max_retries`
      </td>

      <td>
        `sandbox_max_attempts`
      </td>
    </tr>

    <tr>
      <td>
        `scicode_tool_use`
      </td>

      <td>
        `sandbox_retry_delay_seconds`
      </td>

      <td>
        `sandbox_retry_interval_seconds`
      </td>
    </tr>

    <tr>
      <td>
        `scicode_tool_use`
      </td>

      <td>
        `sandbox_api_timeout_seconds`
      </td>

      <td>
        `sandbox_request_timeout_seconds`
      </td>
    </tr>

    <tr>
      <td>
        `stirrup`
      </td>

      <td>
        `stirrup_version`
      </td>

      <td>
        `version`
      </td>
    </tr>

    <tr>
      <td>
        `stirrup`
      </td>

      <td>
        `request_timeout`
      </td>

      <td>
        `request_timeout_seconds`
      </td>
    </tr>

    <tr>
      <td>
        `terminus2`
      </td>

      <td>
        `service_url`
      </td>

      <td>
        `endpoint`
      </td>
    </tr>

    <tr>
      <td>
        `terminus2`
      </td>

      <td>
        `url`
      </td>

      <td>
        `endpoint`
      </td>
    </tr>

    <tr>
      <td>
        `terminus2`
      </td>

      <td>
        `service_headers`
      </td>

      <td>
        `headers`
      </td>
    </tr>

    <tr>
      <td>
        `terminus2`
      </td>

      <td>
        `request_timeout`
      </td>

      <td>
        `request_timeout_seconds`
      </td>
    </tr>
  </tbody>
</table>

OpenHands `max_iterations` limits each run, and SciCode `tool_use_max_loops` limits each substep. Their shared `max_turns` parameter sets an additional task-wide limit.

A few aliases preserve an earlier value convention:

* In `openai_chat`, the Harness alias `max_retries` counts total attempts, including the first. The SDK option in `model.params` counts additional retries.
* In `openevolve`, `iterations=0` selects `baseline_only=true`; positive values set `max_turns`.
* In `scicode_tool_use`, `sandbox_api_timeout_seconds` is an extra HTTP allowance. The total `sandbox_request_timeout_seconds` is `2 * tool_timeout_seconds + sandbox_api_timeout_seconds`.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.