Use OpenMake by task, not by internal subsystem.

Start in chat, then use the sidebar and settings to move between model controls, tasks, research, tools, and administration. Interface language supports Korean, English, Japanese, and Simplified Chinese.

• Choose a model role before a high-cost task
• Use the task list to inspect long-running work
• Keep data controls in Settings under the account that owns the workspace

The default local chat route uses qwen3.6-35b-a3b. Model roles let operators choose suitable routes for chat, research, image work, or agent tasks without exposing every credential to every user.

Agent tasks can use the same workspace context with selected tools and specialists. Skill auto-selection is available to chat and tasks only when SKILL_AUTO_SELECT_ENABLED=true.

• 100 built-in specialists across 18 domains
• Custom agents and skills extend a controlled catalog
• Desktop local work scopes access to a chosen local folder

Deep Research combines configured skill knowledge and MCP evidence in its research pipeline. NotebookLM grounding is for source-bound questions where the supplied material should constrain the answer.

• Review cited tool and connector evidence
• Use grounding when source fidelity matters
• Keep research scope explicit before starting a long task

OpenMake includes 22 built-in MCP tools and can connect governed external servers. Artifacts make a result easier to inspect and hand off than a long chat reply.

Control language, memory, provider keys, usage, roles, audit, and service settings through the appropriate account and operator surfaces. External BYOK provider usage is exempt from the local token quota in 1.13.0; the ChatGPT OAuth role path records usage.

• Use API keys with a role and rate policy
• Review audit records for privileged actions
• Monitor migrations, model routes, connectors, and resource health