Select a checkpoint
Choose a checkpoint from the list to view its state, tool call history, and diff.
AgentTime snapshots your agent's full state — context, memory, tool calls, decisions — continuously. When it goes off-track, roll back to any checkpoint in one click.
from agenttime import AgentTime agent = AgentTime("procurement-agent", auto_snapshot=True) # Auto-snapshots every 5 minutes agent.snapshot("Before the big RFP decision") # Something went wrong? Roll back. state = agent.rollback("ckpt_abc123")
A single bad LLM call can corrupt your agent's memory and send it into a decision spiral. Restarting means losing all progress.
When a database has a bad write, you revert. When your agent makes 50 bad decisions, you're starting from scratch. There's no undo.
Agents running procurement, finance, or legal workflows can't afford to lose state. Yet there's no operational resilience layer for agents.
Every N minutes or every K tool calls, AgentTime captures your agent's full state — context, memory, tool history, decisions.
Visual checkpoint history with diff views. See exactly what changed between any two checkpoints.
Something went wrong? Restore your agent to any past checkpoint instantly. No restart. No lost progress.
from agenttime.integrations.langchain import AgentTimeCallbackHandler from langchain.agents import AgentExecutor, create_react_agent handler = AgentTimeCallbackHandler( agent_id="procurement-agent", auto_snapshot=True, snapshot_interval=300, snapshot_on_tool_call=True, ) executor = AgentExecutor( agent=agent, tools=tools, callbacks=[handler], )
Git for your AI agent's brain. Snapshot, browse, and roll back — in one line of code.