Claude Code Viewer
View and share Claude Code sessions without leaking API keys, your name, email, or local file paths.
Sessions are parsed in your browser — nothing uploads.
npx mesmertools-session-bridgeThe bridge lists your Claude Code sessions by title, only on 127.0.0.1 — nothing leaves your machine. Stop it anytime with Ctrl+C.
How it works
Turn a raw Claude Code export into a safe, readable link in three steps.
Drop the session
Choose a .jsonl file or paste its contents. Parsing happens entirely in your browser.
Review auto-redaction
Sensitive values become labeled chips. Tune categories and add custom terms live.
Share the clean copy
Create an expiring link containing only the normalized, hard-redacted document.
Why use this viewer
Made for real agent sessions, where privacy and navigation matter equally.
Privacy-first redaction
Secrets, identity details, emails, public IPs, and your own terms are masked locally.
Expiring links
Links expire exactly when you choose, with a private delete token.
Readable agent activity
Tool calls, thinking, results, subagents, and prose get the hierarchy long sessions need.
Offline until you save
Files never upload during parsing or redaction. Saving is a separate, explicit action.
Common use cases
Share what the agent did without pasting an unreadable terminal dump.
Bug reports and PRs
Attach the exact debugging path, commands, and results while keeping local paths and credentials out of the thread.
Used by open-source maintainers and contributors.
Team debugging
Give teammates a navigable account of an agent run so they can inspect decisions and tool output asynchronously.
Used by AI engineers and tech leads.
Technical writing
Turn an interesting agent run into source material for a blog post, tutorial, or transparent build log.
Used by devrels and technical writers.
Perfect for
Anyone who needs Claude Code sessions to be understandable outside their own terminal.
Compare long agent runs and share exact implementation context safely.
Agent builders, platform engineers
Publish convincing tutorials with the real interaction preserved.
Devrels, technical advocates
Review how a task unfolded without replaying the entire run.
Engineering managers, staff engineers
Receive useful session evidence without asking contributors to expose their machines.
Maintainers, reviewers