Skip to content
SDKs
PY sdk-python TS sdk-typescript
Organizations
strands-agents strands-labs

strands.types.agent

Agent-related type definitions for the SDK.

This module defines the types used for an Agent.

Limits

Section titled “Limits”
class Limits(TypedDict)

Defined in: src/strands/types/agent.py:17

Per-invocation budget caps for the agent loop.

Each cap, when set, bounds a single invoke_async / stream_async call only; counters are not cumulative across reuses of the same agent. Caps are checked at the top of each loop iteration, so tools requested by the previous turn always run to completion before a cap fires and agent.messages remains in a reinvokable state.

Each cap, when set, must be a positive int. Omit any field (or pass limits=None) for no limit on that dimension.

Priority on simultaneous trip (highest first): turns, total_tokens, output_tokens. The corresponding stop_reason is "limit_turns", "limit_total_tokens", or "limit_output_tokens".

Attributes:

  • turns - Maximum number of agent loop iterations (turns). One turn is one model call plus any tool execution that follows. Counted against len(metrics.latest_agent_invocation.cycles).
  • output_tokens - Maximum cumulative model-generated tokens, summed across every model call in the loop (metrics.latest_agent_invocation.usage["outputTokens"]). Distinct from per-call provider-level caps, which bound a single model call’s output. Soft cap: a single oversized response can overshoot the budget; checked at turn boundaries, not within an individual model call.
  • total_tokens - Maximum cumulative input + output tokens (metrics.latest_agent_invocation.usage["totalTokens"]). Each model call’s input includes prior turns, so this counter compounds across the run and approximates total token spend. Soft cap, same caveat as output_tokens.

ConcurrentInvocationMode

Section titled “ConcurrentInvocationMode”
class ConcurrentInvocationMode(str, Enum)

Defined in: src/strands/types/agent.py:52

Mode controlling concurrent invocation behavior.

Values: THROW: Raises ConcurrencyException if concurrent invocation is attempted (default). UNSAFE_REENTRANT: Allows concurrent invocations without locking.

Warnings:

The UNSAFE_REENTRANT mode makes no guarantees about resulting behavior and is provided only for advanced use cases where the caller understands the risks.