Runtime

The runtime is the journaling layer. Every model call and every tool dispatch is keyed by (session_id, step_name) and stored. On a fresh process pointed at the same journal, the same session resumes , already-completed steps return cached results; only the un-completed work re-executes.

from loomflow import Agent
from loomflow.runtime import SqliteRuntime
 
agent = Agent(
    "...",
    model="claude-opus-4-7",
    runtime=SqliteRuntime("./journal.db"),
)

Three backends, same protocol

What gets journaled

Every wrap point in the agent loop:

• model_call_<turn>. The streamed model response (text + tool calls + usage).
• tool_call_<turn>_<slot>. Each parallel tool dispatch in a turn.
• persist_episode_<turns>. The memory.remember(episode) call.

Architecture-specific steps add more, delegate_<turn>_<worker> for Supervisor, plan / step_N / synthesize for Plan-and-Execute, etc.

Default, InProcRuntime

Without a runtime= kwarg, Agent uses InProcRuntime. No journal, no durability. Crashes lose everything. Fine for stateless one-shot scripts and tests.

When to upgrade

Where journaling matters

• Long-running runs. Research agents, multi-stage pipelines , one model call gone wrong shouldn’t lose the work.
• Expensive tool calls. Slow APIs, paid endpoints, work that’s hard to redo.
• Resumability after deploys. Restart the process, pick up ongoing runs.