Built-in tools

The Tase model ships with native awareness of the tools that power the Tase assistant. Their schemas — names, parameters, and when each one applies — are trained directly into the model’s weights. That means a request can trigger a task, health, or finance tool call without your request ever describing those tools.

This is the key difference from custom tools. Custom tools travel in the tools array of every request, as described in Function calling. Built-in tools don’t have to — the model already knows them. If you do send a schema for a built-in tool, the request is accepted and works normally; it is simply unnecessary.

A built-in call on the wire

The wire contract is identical to custom tools: the model returns tool_calls, your application executes and replies with role: "tool" messages. Note that the request below sends no tools array at all.

{
  "model": "tase-8b",
  "messages": [
    { "role": "user", "content": "Add a task to renew my passport next Friday." }
  ]
}

{
  "choices": [
    {
      "index": 0,
      "finish_reason": "tool_calls",
      "message": {
        "role": "assistant",
        "content": null,
        "tool_calls": [
          {
            "id": "call_7h2h",
            "type": "function",
            "function": {
              "name": "create_task",
              "arguments": "{\"title\": \"Renew passport\", \"due_date\": \"2026-07-24\"}"
            }
          }
        ]
      }
    }
  ]
}

The tool families

The full catalog is grouped into seven families. Rather than memorizing individual tool names, think in terms of what each family covers — the model maps natural language onto the right tool for you.

Tasks and day planning

Create, reschedule, complete, and query tasks; lay out the day as time blocks; manage quests and rewards. This is the family most requests end up touching.

Move my gym session to 6 pm and add a packing task before it.Example user request

Health tracking

Log meals and calories, water, medications, supplements, workouts, vitals, and habits — and read them back as summaries and trends.

Log two eggs and a coffee for breakfast, and remind me about my vitamin D.Example user request

Finance

Record expenses and income, manage budgets, and answer spending questions against the user’s own numbers instead of guessing.

I spent $40 on groceries — how am I doing against this month’s food budget?Example user request

Memory and AI Brain

Save and recall notes, ideas, knowledge, and long-term memories. This is what makes “say it once, it’s logged” work: a passing remark becomes a durable record the model can retrieve later.

Remember that Deniz prefers morning meetings.Example user request

Agents and automation

Spin up background agents, schedule recurring routines, and check on running jobs — work that continues after the conversation ends.

Every Sunday evening, prepare a summary of my week and draft Monday’s plan.Example user request

Connections: calendar, email, and files

Read and manage calendar events, search and draft email, and find or fetch files from connected accounts.

What’s on my calendar tomorrow, and did the contract PDF arrive in my inbox?Example user request

World knowledge: weather, places, search, and vision

Look up current weather, find places nearby, search the web, and understand images — the family that grounds answers in the world outside the user’s own data.

Find a quiet coffee shop near me and check if it will rain this afternoon.Example user request

Restricting and overriding

Everything from Function calling applies to built-ins too. tool_choice: "none" disables all tool calling for a request; the named form forces one specific tool. And because sending a schema for a built-in tool is accepted, you can redeclare one when you want to pin its exact shape for your integration.