Agent Trace Mechanism

Agent Trace Mechanism is an open source history reader and tool set that lets AI agents read local Codex and Claude Code conversations.

Agent Trace reads session files that Codex and Claude Code save locally on a computer. An AI agent in either app can read chats from both sources, including messages and recorded tool activity from running chats. The Model Context Protocol server runs locally and provides tools for finding and reading histories. The package includes the underlying Python library, a desktop app preview, and a standalone MCP server. It supports compact and detailed views, search functions for Codex operations, and subagent inclusion. The project uses JSONL session files and a local JSON manifest configuration file for privacy.

Agent Trace Mechanism at a glance

Best for: Software engineer
Runs on: macOS, Linux, API, self-hosted, open source
Pricing: Free and open source
Released: November 2026
Category: AI coding tool
Website: github.com/wcook04/agent-trace-mechanism

Key features

Read Codex and Claude Code session files
Run a local Model Context Protocol server
Choose session files via a manifest
View compact transcripts and detailed conversations
Search commands, outputs, and patches
Compare running chat reads for changes
Include subagents in queries

Use it for

Read earlier Codex and Claude Code conversations
Check failed tests, corrections, and retries
Search commands, outputs, and patches in Codex
Compare changes in running chats

Companions

Tools that articles about Agent Trace Mechanism recommend using with it, each with its job.

Codex — Built to work with Codex.
Claude Code — Built to work with Claude Code.
Context7 — Provides up-to-date documentation for AI agents and code editors.
Superpowers — Plan, test and execute workflows across harnesses.
Firecrawl — Provides direct web data access for AI agents with open source flexibility.
Composio — One login for GitHub, Linear, Slack and 1,000+ apps.

Alternatives to Agent Trace Mechanism

Agent Archive — Hand off coding sessions between Claude Code, Codex, and Cursor across machines.
claude-mem — It shares persistent, searchable agent memory across sessions and devices.
HarnessRouter — One API to manage agent runtimes, sandboxes, and execution lifecycles.
Claude Tap — Inspect prompts, tool calls, and token usage locally without a hosted dashboard.
Claude Codex Bridge — Runs multiple AI coding agents concurrently with stable cross-provider collaboration.
Reviewer Agent — It uses a separate AI model to review code and prevent premature agent stopping.
Codex Host — Runs different AI coding assistants simultaneously in the native Codex Desktop window.
Ret — Manage multiple AI coding agents across projects from one unified desktop interface.
Runway — Control Claude Code and Codex sessions on local or cloud machines from any phone.
Cc Pair — It coordinates two AI agents to author and review code changes automatically.
Claude Code Router — It centralizes control of AI agent requests with intelligent routing and observability.
Claude Super Sidebar — It brings agent relationships, execution states, and task flows into one panel.

All Agent Trace Mechanism alternatives