Thinking steps
The reasoning Claude does before each action. The reasoning the terminal hides — and where most of the surprising decisions actually live.
hidden by default
Your Claude is coding blind.
See everything it did — every file path, tool call, and token — from the session logs already on your machine.
Used by 50,000+ developers·Open source·No API keys
The problem
Bring back the visibility Claude took away.
Since v2.1.20, detailed output was replaced with opaque summaries. The community backlash was immediate. The terminal hides it — claude-devtools puts it back.
Terminalwhat Claude shows you
claude-devtoolswhat actually happened
What the terminal strips out.
Six categories of detail Claude Code stopped showing. All of it is still in ~/.claude/ — just not on your screen.
Tool call I/O
Exact inputs and outputs for every Read, Edit, Bash, Grep.
Subagent activity
Per-agent tool traces with tokens, duration, and cost.
Context window breakdown
What filled it — CLAUDE.md, @-mentions, skills, tool output.
Team coordination
How sub-agents handed off work and what each one saw.
File paths & content
The actual files touched and the diffs that landed.
All invisible in the terminal. The only built-in option is --verbose, which dumps thousands of lines of raw JSON noise.
Not a wrapper. Doesn't touch Claude Code.
Reads ~/.claude/ — the session logs already on your machine. Works with every session you've ever run, from any tool.
Key Features
01
Context Reconstruction
Per-turn token attribution across 7 categories — CLAUDE.md (global, project, directory), skill activations, @-mentioned files, tool I/O, thinking, team overhead, user text. Visualizes how your context fills, compresses during compaction, and refills — so you can see exactly what was in the window at any point.
02
Copy & Paste Without the Terminal Pain
Copying from the terminal is awkward — selection wraps, ANSI codes leak, multi-line breaks. claude-devtools renders every message, tool call, and output as real selectable text. One click to copy a code block, file path, or entire response. Works the same on macOS, Linux, and Windows.
03
Tool Call Inspector
Syntax-highlighted Read calls, inline Edit diffs, Bash output, and full subagent execution trees. Every collapsed "Read 3 files" expanded.
04
Project Memory
Per-project Claude memory lives at ~/.claude/projects/<project>/memory/ — MEMORY.md as an index plus one .md per layer (working style, architecture notes, etc.). Opens in a dedicated pane: layer list on the left, full markdown rendering with in-pane search on the right, and an "Open in…" launcher that hands the file off to Finder, Cursor, VS Code, Zed, Xcode, iTerm, Ghostty, Terminal, or copies the absolute path.
05
Team & Subagent Trees
Isolated execution trees per agent with their own tool traces, token metrics, duration, and cost. Nested agents render as recursive trees.
06
Notification Triggers
System notifications with built-in alerts for .env file access, tool execution errors, and high token usage. Add regex triggers for sensitive paths (for example payment/billing files), match specific tool fields, set input/output/total token thresholds, and filter by trigger color.
07
SSH Remote Sessions
Inspect sessions on any remote machine over SSH. Reads ~/.ssh/config, supports agent forwarding and key auth. Switch between local and remote instantly.
08
Command Palette
Cmd+K opens cross-session search. Results show context snippets with highlighted keywords. Navigate directly to the exact message.
09
Multi-Pane Layout
Open multiple sessions side-by-side. Drag-and-drop tabs between panes, split views, compare sessions in parallel.
installOne line. Then open every session
One line. Then open every session
you've ever recorded.
bash
brew install --cask claude-devtoolsHomebrew on macOS. Direct downloads below for other platforms.
Open source · MITNo login · no API keysZero outbound callsWorks on every session already on disk