feat(checkpoint): wire checkpointing into agent event loop

src/strands/agent/agent.py

@@ -29,6 +29,7 @@
 from .._async import run_async
 from ..event_loop._retry import ModelRetryStrategy
 from ..event_loop.event_loop import INITIAL_DELAY, MAX_ATTEMPTS, MAX_DELAY, event_loop_cycle
+from ..experimental.checkpoint import Checkpoint
 from ..tools._tool_helpers import generate_missing_tool_result_content
 from ..types._snapshot import (
     SNAPSHOT_SCHEMA_VERSION,
@@ -146,6 +147,7 @@ def __init__(
         tool_executor: ToolExecutor | None = None,
         retry_strategy: ModelRetryStrategy | _DefaultRetryStrategySentinel | None = _DEFAULT_RETRY_STRATEGY,
         concurrent_invocation_mode: ConcurrentInvocationMode = ConcurrentInvocationMode.THROW,
+        checkpointing: bool = False,
     ):
         """Initialize the Agent with the specified configuration.
 
@@ -214,6 +216,11 @@ def __init__(
                 Set to "unsafe_reentrant" to skip lock acquisition entirely, allowing concurrent invocations.
                 Warning: "unsafe_reentrant" makes no guarantees about resulting behavior and is provided
                 only for advanced use cases where the caller understands the risks.
+            checkpointing: When True, the event loop pauses at cycle boundaries (after model call,
+                after all tools execute) and returns an AgentResult with stop_reason="checkpoint"
+                and a populated ``checkpoint`` field. Persist the checkpoint and resume by passing a
+                ``CheckpointResumeContent`` block as the next prompt. Defaults to False.
+                See :mod:`strands.experimental.checkpoint` for usage and limitations.
 
         Raises:
             ValueError: If agent id contains path separators.
@@ -304,6 +311,10 @@ def __init__(
 
         self._interrupt_state = _InterruptState()
 
+        # Checkpointing: when True, event loop pauses at cycle boundaries
+        self._checkpointing: bool = checkpointing
+        self._checkpoint_resume_context: Checkpoint | None = None
+
         # Runtime state for model providers (e.g., server-side response ids)
         self._model_state: dict[str, Any] = {}
 
@@ -998,6 +1009,46 @@ async def _convert_prompt_to_messages(self, prompt: AgentInput) -> Messages:
         if self._interrupt_state.activated:
             return []
 
+        # Resume detection — must run before existing shape handling so checkpointResume
+        # blocks aren't misinterpreted as content blocks. Mirrors _InterruptState.resume()
+        # conventions (TypeError for shape, KeyError for lookup, ValueError for misconfig;
+        # error messages use the SDK's key=<value> | message format).
+        if isinstance(prompt, list) and prompt:
+            has_checkpoint_resume = any(isinstance(c, dict) and "checkpointResume" in c for c in prompt)
+            if has_checkpoint_resume:
+                if not self._checkpointing:
+                    raise ValueError(
+                        "Received checkpointResume block but agent was created with "
+                        "checkpointing=False. Pass checkpointing=True when constructing "
+                        "the Agent to enable durable execution."
+                    )
+
+                invalid_types = [
+                    key
+                    for content in prompt
+                    if isinstance(content, dict)
+                    for key in content
+                    if key != "checkpointResume"
+                ]
+                if invalid_types:
+                    raise TypeError(
+                        f"content_types=<{invalid_types}> | checkpointResume cannot be mixed with other content types"
+                    )
+
+                if len(prompt) != 1:
+                    raise TypeError(
+                        f"block_count=<{len(prompt)}> | only one checkpointResume block permitted per prompt"
+                    )
+
+                resume_block = prompt[0].get("checkpointResume", {})
+                if not isinstance(resume_block, dict) or "checkpoint" not in resume_block:
+                    raise KeyError("checkpoint | missing required key in checkpointResume block")
+
+                checkpoint = Checkpoint.from_dict(resume_block["checkpoint"])
+                self.load_snapshot(Snapshot.from_dict(checkpoint.snapshot))
+                self._checkpoint_resume_context = checkpoint
+                return []
+
         messages: Messages | None = None
         if prompt is not None:
             # Check if the latest message is toolUse

src/strands/agent/agent_result.py

@@ -9,6 +9,7 @@
 
 from pydantic import BaseModel
 
+from ..experimental.checkpoint import Checkpoint
 from ..interrupt import Interrupt
 from ..telemetry.metrics import EventLoopMetrics
 from ..types.content import Message
@@ -26,6 +27,9 @@ class AgentResult:
         state: Additional state information from the event loop.
         interrupts: List of interrupts if raised by user.
         structured_output: Parsed structured output when structured_output_model was specified.
+        checkpoint: Checkpoint captured when the agent paused for durable execution.
+            Populated only when stop_reason == "checkpoint". See
+            strands.experimental.checkpoint for usage.
     """
 
     stop_reason: StopReason
@@ -34,6 +38,7 @@ class AgentResult:
     state: Any
     interrupts: Sequence[Interrupt] | None = None
     structured_output: BaseModel | None = None
+    checkpoint: Checkpoint | None = None
 
     @property
     def context_size(self) -> int | None:
@@ -85,15 +90,23 @@ def from_dict(cls, data: dict[str, Any]) -> "AgentResult":
         Returns:
             AgentResult instance
         Raises:
-            TypeError: If the data format is invalid@
+            TypeError: If the data format is invalid
         """
         if data.get("type") != "agent_result":
             raise TypeError(f"AgentResult.from_dict: unexpected type {data.get('type')!r}")
 
         message = cast(Message, data.get("message"))
         stop_reason = cast(StopReason, data.get("stop_reason"))
+        checkpoint_data = data.get("checkpoint")
+        checkpoint = Checkpoint.from_dict(checkpoint_data) if checkpoint_data else None
 
-        return cls(message=message, stop_reason=stop_reason, metrics=EventLoopMetrics(), state={})
+        return cls(
+            message=message,
+            stop_reason=stop_reason,
+            metrics=EventLoopMetrics(),
+            state={},
+            checkpoint=checkpoint,
+        )
 
     def to_dict(self) -> dict[str, Any]:
         """Convert this AgentResult to JSON-serializable dictionary.
@@ -105,4 +118,5 @@ def to_dict(self) -> dict[str, Any]:
             "type": "agent_result",
             "message": self.message,
             "stop_reason": self.stop_reason,
+            "checkpoint": self.checkpoint.to_dict() if self.checkpoint else None,
         }

src/strands/event_loop/event_loop.py

@@ -15,6 +15,7 @@
 
 from opentelemetry import trace as trace_api
 
+from ..experimental.checkpoint import Checkpoint
 from ..hooks import AfterModelCallEvent, BeforeModelCallEvent, MessageAddedEvent
 from ..telemetry.metrics import Trace
 from ..telemetry.tracer import Tracer, get_tracer
@@ -122,6 +123,19 @@ async def event_loop_cycle(
     # Initialize state and get cycle trace
     if "request_state" not in invocation_state:
         invocation_state["request_state"] = {}
+
+    # Consume checkpoint resume context (one-shot: cleared after reading).
+    # Cross-invocation state (resume context) lives on the agent; within-cycle
+    # transient state (resume position for the skip check, cycle index) lives
+    # in invocation_state.
+    resume_ctx = agent._checkpoint_resume_context
+    if resume_ctx is not None:
+        agent._checkpoint_resume_context = None
+        # after_tools completed that cycle, so next cycle starts at +1
+        next_cycle = resume_ctx.cycle_index + 1 if resume_ctx.position == "after_tools" else resume_ctx.cycle_index
+        invocation_state["_checkpoint_cycle_index"] = next_cycle
+        invocation_state["_checkpoint_resume_position"] = resume_ctx.position
+
     attributes = {"event_loop_cycle_id": str(invocation_state.get("event_loop_cycle_id"))}
     cycle_start_time, cycle_trace = agent.event_loop_metrics.start_cycle(attributes=attributes)
     invocation_state["event_loop_cycle_trace"] = cycle_trace
@@ -181,6 +195,32 @@ async def event_loop_cycle(
                 )
 
             if stop_reason == "tool_use":
+                # Checkpoint after model call, before tool execution.
+                # One-shot pop: safe because after_model always returns before reaching
+                # after_tools, so the stashed position is only consumed once.
+                if agent._checkpointing:
+                    resume_pos = invocation_state.pop("_checkpoint_resume_position", None)
+                    if resume_pos == "after_model":
+                        pass  # Just resumed here — skip re-checkpoint, proceed to tools
+                    else:
+                        cycle_index = invocation_state.get("_checkpoint_cycle_index", 0)
+                        checkpoint = Checkpoint(
+                            position="after_model",
+                            cycle_index=cycle_index,
+                            snapshot=agent.take_snapshot(preset="session").to_dict(),
+                        )
+                        agent.event_loop_metrics.end_cycle(cycle_start_time, cycle_trace)
+                        if cycle_span:
+                            tracer.end_event_loop_cycle_span(span=cycle_span, message=message)
+                        yield EventLoopStopEvent(
+                            "checkpoint",
+                            message,
+                            agent.event_loop_metrics,
+                            invocation_state["request_state"],
+                            checkpoint=checkpoint,
+                        )
+                        return
+
                 # Handle tool execution
                 tool_events = _handle_tool_execution(
                     stop_reason,
@@ -590,6 +630,24 @@ async def _handle_tool_execution(
         )
         return
 
+    # Checkpoint after all tools complete, before the next model call.
+    if agent._checkpointing:
+        cycle_index = invocation_state.get("_checkpoint_cycle_index", 0)
+        invocation_state["_checkpoint_cycle_index"] = cycle_index + 1
+        checkpoint = Checkpoint(
+            position="after_tools",
+            cycle_index=cycle_index,
+            snapshot=agent.take_snapshot(preset="session").to_dict(),
+        )
+        yield EventLoopStopEvent(
+            "checkpoint",
+            message,
+            agent.event_loop_metrics,
+            invocation_state["request_state"],
+            checkpoint=checkpoint,
+        )
+        return
+
     events = recurse_event_loop(
         agent=agent, invocation_state=invocation_state, structured_output_context=structured_output_context
     )

src/strands/experimental/checkpoint/__init__.py

@@ -8,5 +8,12 @@
 """
 
 from .checkpoint import CHECKPOINT_SCHEMA_VERSION, Checkpoint, CheckpointPosition
+from .types import CheckpointResumeContent, CheckpointResumeDict
 
-__all__ = ["CHECKPOINT_SCHEMA_VERSION", "Checkpoint", "CheckpointPosition"]
+__all__ = [
+    "CHECKPOINT_SCHEMA_VERSION",
+    "Checkpoint",
+    "CheckpointPosition",
+    "CheckpointResumeContent",
+    "CheckpointResumeDict",
+]

src/strands/experimental/checkpoint/checkpoint.py

@@ -17,6 +17,14 @@
 - State via AgentResult.checkpoint field
 - Resume via checkpointResume content block in next agent() call
 
+Interaction with interrupts:
+- Interrupts take priority over checkpoints. If a tool raises an Interrupt
+  during a checkpointing=True cycle, the event loop returns
+  stop_reason="interrupt" (not "checkpoint"). The after_tools checkpoint
+  is never reached because the interrupt path returns early.
+- This is intentional: interrupts require human input, checkpoints are
+  for worker-level durability. Different semantics, different priorities.
+
 V0 Known Limitations:
 - Metrics reset on each resume call. The caller is responsible for aggregating
   metrics across a durable run. EventLoopMetrics reflects only the current call.
@@ -35,6 +43,8 @@
 from dataclasses import asdict, dataclass, field
 from typing import Any, Literal
 
+from ...types.exceptions import CheckpointException
+
 logger = logging.getLogger(__name__)
 
 CHECKPOINT_SCHEMA_VERSION = "1.0"
@@ -79,11 +89,11 @@ def from_dict(cls, data: dict[str, Any]) -> "Checkpoint":
             data: Serialized checkpoint data.
 
         Raises:
-            ValueError: If schema_version doesn't match the current version.
+            CheckpointException: If schema_version doesn't match the current version.
         """
         version = data.get("schema_version", "")
         if version != CHECKPOINT_SCHEMA_VERSION:
-            raise ValueError(
+            raise CheckpointException(
                 f"Checkpoints with schema version {version!r} are not compatible "
                 f"with current version {CHECKPOINT_SCHEMA_VERSION}."
             )

src/strands/experimental/checkpoint/types.py

@@ -0,0 +1,37 @@
+"""Content-block types for checkpoint resume.
+
+Mirrors the interrupt pattern (`InterruptResponseContent` in `types/interrupt.py`).
+Stays under `experimental/checkpoint/` for V0; will graduate to
+`src/strands/types/checkpoint.py` when the feature exits experimental.
+"""
+
+from typing import Any, TypedDict
+
+
+class CheckpointResumeDict(TypedDict):
+    """Inner payload for a checkpointResume content block.
+
+    Attributes:
+        checkpoint: Serialized Checkpoint as produced by ``Checkpoint.to_dict()``.
+    """
+
+    checkpoint: dict[str, Any]
+
+
+class CheckpointResumeContent(TypedDict):
+    """Content block that resumes a paused durable agent.
+
+    Pass a list containing exactly one instance of this type as the prompt to
+    ``Agent.invoke_async`` / ``Agent.__call__`` to resume from a checkpoint.
+
+    Example::
+
+        result = await agent.invoke_async(
+            [{"checkpointResume": {"checkpoint": previous_checkpoint.to_dict()}}]
+        )
+
+    Attributes:
+        checkpointResume: The resume payload carrying the serialized checkpoint.
+    """
+
+    checkpointResume: CheckpointResumeDict

src/strands/types/_events.py

@@ -22,6 +22,7 @@
 if TYPE_CHECKING:
     from ..agent import AgentResult
     from ..agent._agent_as_tool import _AgentAsTool
+    from ..experimental.checkpoint import Checkpoint
     from ..multiagent.base import MultiAgentResult, NodeResult
 
 
@@ -227,6 +228,7 @@ def __init__(
         request_state: Any,
         interrupts: Sequence[Interrupt] | None = None,
         structured_output: BaseModel | None = None,
+        checkpoint: "Checkpoint | None" = None,
     ) -> None:
         """Initialize with the final execution results.
 
@@ -237,8 +239,11 @@ def __init__(
             request_state: Final state of the agent execution
             interrupts: Interrupts raised by user during agent execution.
             structured_output: Optional structured output result
+            checkpoint: Optional checkpoint when stop_reason == "checkpoint".
         """
-        super().__init__({"stop": (stop_reason, message, metrics, request_state, interrupts, structured_output)})
+        super().__init__(
+            {"stop": (stop_reason, message, metrics, request_state, interrupts, structured_output, checkpoint)}
+        )
 
     @property
     @override

src/strands/types/exceptions.py

@@ -111,3 +111,9 @@ class ConcurrencyException(Exception):
     """
 
     pass
+
+
+class CheckpointException(Exception):
+    """Exception raised when checkpoint operations fail (e.g., incompatible schema version)."""
+
+    pass