Building AgentTrace: How I Made AI Agents Remember, Recover, and Resume with FalkorDB

Description and use

AgentTrace uses FalkorDB to store and manage AI agents’ execution state, enabling them to remember, recover, and resume after failures.

Problem

AI agents forget what they have already done, making it hard to resume or recover after crashes or failures.

Outcome

AgentTrace successfully records failures, introduces repair tasks, updates dependencies, and verifies outcomes, allowing agents to resume work without starting from scratch.

Workflow

• 1. Planner creates a validated task plan with dependencies.
• 2. Executor selects eligible tasks, claims them, runs tools, and records results.
• 3. Verifier independently checks task outcomes.
• 4. Recovery investigates failures, retrieves graph evidence, and introduces corrective actions.
• 5. All roles use the shared FalkorDB graph to enforce dependencies and prevent duplicate task claims.

Evidence

AI agents are getting better at planning, writing code, and using tools. But there is still a frustrating problem: what happens when an agent forgets what it has already done?

Source

AgentTrace records execution state inside a persistent graph.

Source

FalkorDB answers these questions through graph queries.

Source

The demo includes a controlled Python calculator application with a deliberately broken addition function.

Source