Mindwalk turns AI coding-agent logs into interactive 3D light shows on a map of your codebase

A new open-source tool called Mindwalk lets developers replay AI coding-agent sessions as animated 3D visualizations, turning dry JSONL logs into a luminous night map of the repository. The tool, now available on GitHub, aims to make the decision-making process of automated coding systems transparent and auditable, according to mlllm.io.

Instead of scrolling through raw session logs that don't convey how an agent interpreted a task, Mindwalk shows the codebase as a dark map and agent activity as light trails moving through it. Hacker News reported that the tool visualizes "which parts of the repo it treated as relevant, where it explored, and whether its footprint matched expectations."

"The Mindwalk tool allows for the playback of AI agent sessions on a 3D map of the codebase using 'light trails.'"

That light-trail approach turns searches, reads, and edits into streaks of illumination that brighten areas of interest, leaving untouched parts of the repository dark. Hacker News described the effect as "a night map, with agent activity shown as light moving through the codebase." The contrast helps developers quickly assess whether the agent's actual path aligns with what they expected.

"Observability tools for AI agents make the decision-making process of automated systems transparent and auditable."

Mindwalk addresses a growing need as more teams rely on autonomous coding agents in production. By making agent behavior visible, the tool helps developers debug unexpected actions, verify that the agent explored the right files, and build trust in automated workflows. Both outlets emphasized that the tool is open-source, inviting the community to adapt it to their own codebase maps and agent frameworks.