Choose theme
AI Resources
Her Claude Code Session Detective
Her is a Hugging Face Space for reading Claude Code session JSONL traces and turning them into a plain-English view of what happened during a coding-agent run.
The Hugging Face blog describes Her as a small hackathon-style trace reader that reconstructs sessions, surfaces production/config/secret-related activity, shows tool and token usage, and answers questions with turn references. Use this as a first read, not a recommendation. Open the original project before trusting details like terms, limits, privacy, cost, setup, or safety.
What it is
Claude Code session trace reader
Her is built around Claude Code session files: the JSONL traces that can contain user turns, tool calls, token usage, subagent activity, skills, MCP servers, and other run details.
Why it stands out
Session story instead of raw JSON
The value is readability. The blog says Her reconstructs what happened in plain English, links findings back to turns, and lets readers ask questions about why a tool or action appeared in the trace.
Availability
Published Space, README and local path
The public Space contains the code and setup notes, including a local ./her option. Its hosted runtime was paused on 4 October 2026; the June 2026 blog describes the project and its trace-reading workflow.
Why it matters
What makes it useful
When a session explanation mentions a surprising tool call, the turn links give the reader a place to check what was actually recorded. The author describes Her as reconstructing the run and answering questions about its actions, so the explanation can lead back to a specific event instead of standing alone as a story.
What to know
Where it fits
Her reads saved Claude Code JSONL files after the run. The README offers both a hosted Space and a local ./her path; it does not control the original coding session. The hosted Space was paused when checked on 4 October 2026, so its published code and local setup are the available starting points while the demo is paused.
Notable points
What stands out
The author separates a deterministic findings engine from Nemotron-Mini-4B-Instruct narration and softer suggestions. A prose answer therefore does not have the same status as a finding linked to a recorded turn, and neither replaces the original trace.
Before using
What to review
Claude Code traces can contain private code, prompts, paths, credentials and customer information. The README describes uploader redaction as best-effort, so it does not replace checking the file before sharing it.
The author documents a 24-hour sweeper and best-effort clearing when a tab closes. Those are handling descriptions, not a guarantee that every copy or secret is removed.
Choose between hosted processing and the documented local setup. Self-hosting the ZeroGPU Space also has its own plan and runtime requirements; a local interface does not by itself establish every model connection is offline.
Follow turn references back to the saved JSONL when an action or conclusion matters. Her is a trace reader, rather than a complete audit or governance system.
Reader fit
Who may find it relevant
Claude Code users who want a quicker first pass over long session traces.
Teams comparing coding-agent observability, token usage, tool usage, and post-run review ideas.
Less relevant for readers who do not use Claude Code or do not have session JSONL traces to inspect.
Editorial note
Why LifeHubber lists it
One JSONL file gives Her a session view; several files can form a project view. The latter is useful when the question spans runs, such as locating which session recorded a particular change. The author describes answers that point back to the relevant session, rather than requiring the reader to search each conversation separately.
Source links
Source materials
Reader note
Before relying on this entry
LifeHubber lists entries to help readers inspect AI projects, not to endorse them or prove they are safe, suitable, accurate, maintained, or right for a specific use. We do not verify every entry in depth. Before relying on anything listed, review the original materials, terms, privacy practices, limits, and risks that matter for your situation.
What to explore next
Continue from one trace to broader session evidence.
Her reconstructs what happened in a Claude Code session. Continue with a local coaching view across several coding harnesses or a focused report of token use, models, sessions, and estimated cost.
More in AI Agents
Keep browsing this category
Explore more AI agent projects.
Paperclip
paperclipai/paperclip
A self-hosted server and dashboard for coordinating agent teams through companies, goals, roles, issues, heartbeats, budgets, approvals, and persistent activity records.
Open Instinct
mariagorskikh/open-instinct
A beta personal-agent project with local CLI chat, messaging and app integrations, readable memory, and permission tiers for requests from other people and agents. Its self-hosted messaging tunnel has a reported delivery problem with a fix proposed.
Gemma 4 Browser Extension
nico-martin/gemma4-browser-extension
An independent Chrome extension experiment for running an on-device browser agent with Transformers.js, WebGPU, Gemma 4, page RAG, tab tools, and semantic history search.
For project maintainers
Listed here? You can use the badge.
If you maintain a project with a current LifeHubber listing, you may add the optional “Listed on LifeHubber AI Resources” badge to its README, docs, or website. No introduction or permission request is needed.
