Mindwalk

Mindwalk is a visualization tool that replays coding-agent sessions by projecting activity onto a 3D map of the codebase for analysis.

Mindwalk is a tool for visualizing and reviewing the activity of coding agents by replaying their sessions on a 3D map of a repository. It renders the repository as a night map, where agent actions like searching, reading, or editing illuminate specific areas, leaving the rest of the code dark. The tool supports session logs from Claude Code, Codex, and pi agents. Users can scrub through a timeline of events, inspect file-specific histories, and view subagent traces. An optional evaluation feature allows users to run a local agent CLI to assess session performance against criteria derived from the original task description. The entire interface runs locally, ensuring that session data remains on the user's machine unless an evaluation is explicitly triggered.

Mindwalk at a glance

Best for: Software engineer
Runs on: Web, CLI, Linux, Windows, macOS
Pricing: Free
Released: July 2026
Category: AI coding tool
Website: github.com/cosmtrek/mindwalk

Key features

3D codebase repository visualization
Replay playback for agent sessions
Support for Claude Code, Codex, and pi session logs
Timeline markers for context, subagents, and user turns
File-specific visit history inspection
Local agent-based session evaluation
Export playback as WebM video files

Use it for

Analyzing agent navigation patterns through a codebase
Reviewing how an agent interpreted a specific coding task
Evaluating session performance using custom criteria
Comparing agent sub-process traces against main session logs

Alternatives to Mindwalk

jevgrep — Reduces coding agent costs by providing targeted, relevant repository context.
Agent Beacon — Retains coding agent knowledge across tools so errors do not repeat.
Entelligence AI — Inspect agent sessions locally without digging through logs or sharing data.
AgentsView — It keeps your AI coding session archive local by default while tracking costs and activity.
CodeGraph — It eliminates file crawling, reducing AI agent tool calls by 88% and cutting costs.
Agent Archive — Hand off coding sessions between Claude Code, Codex, and Cursor across machines.
Invariant — Persistent isolated Linux environments and parallel coding agent experiments.
Raindrop Workshop — Provides real-time observability and self-healing evaluation loops for local AI agents.
Agent Skills — Encodes senior-engineer workflows and quality gates into 24 skills.
Task View — Review AI-generated code as an interactive structural map rather than a flat diff.
HumanLayer — Bridges AI agent sessions and human feedback into a single, cohesive interface.
CodeSight — Reduces AI token usage significantly and maps projects instantly.

All Mindwalk alternatives