Token Savings (Machine-Reported Estimate)

Self-reported by each machine's local Claude Code hook (est_tokens_saved) -- an estimate covering that machine's overall Claude Code usage, not just actions taken in this CCDR Workspace. For CCDR's own exact, recorded numbers, see Token Savings Evidence below -- the two are different measurements and are not expected to match.

Context

Projects

Combined across every CCDR component that reduces context -- Prompt Optimizer, Intent Layers, Handoff, Terminal Optimizer, and Conversation Reset.

By Project

Project	Sessions	Tokens Saved	Avg Reduction

By Machine

Machine	Sessions	Tokens Saved	Avg Reduction

Machines

Whether each machine's local Claude Code hook install matches what CCDR ships. Deployment stays a manual, explicit action you run locally -- CCDR only reports status here, it never writes to your filesystem or updates hooks on its own.

Per-project token savings aren't tracked yet (only per-machine); this shows real checkpoint activity per project as the closest honest proxy.

Project	Checkpoints	Avg Summary Size

Select a project to resume exactly where you left off.

Stage 1 · Engineering Task

Describe the engineering work. CCDR builds the complete mission — scope, context, boundaries and acceptance criteria — automatically.

Examples: Fix Avg Delivery Time calculation Investigate Shopify sync mismatch Add Executive Currency Warning Implement SKU drill-down Create Customer Segmentation API

Stage 2 · CCDR Analysis

• ✓ Understanding the Request
• ✓ Defining Scope & Boundaries
• ✓ Loading Project Context
• ✓ Verifying Engineering Rules
• ✓ Mission Ready for Dispatch

This reflects what CCDR already computed for this request in one server round trip — not a live progress stream.

Objective

—

Business Goal

—

Scope

Acceptance Summary

Stage 3 · Dispatch to Claude

Dispatch to Claude

CCDR cannot dispatch Claude directly from this browser. Claude Transport runs the Claude CLI as a local process against your actual working tree — the control plane has no access to your machine. Copy the command below and run it in your project directory; knowledge files are verified and (re)generated automatically the moment you do.

Stage 4 · Mission Activity

Near real-time event streaming via Server-Sent Events, published by the local agent as it actually observes each stage — not a simulated progress bar. Core events (received/started/completed/failed) always appear; the rest appear only when this dispatch could observe them and are simply omitted otherwise.

Stage 5 · Completion Report

Completion Report

Summary

—

Files Touched (from mission scope)

Tests Executed / Passed

Not captured by the current completion schema (status/summary/tasks/decisions/open_issues only).

Warnings / Errors

Not captured as a distinct field — check Open Issues below and the raw summary.

Stage 6 · CCDR Review

CCDR Review

Decisions Recorded

Open Issues / Remaining Risks

ADR updates are not distinguished from Decisions in the current schema.

Stage 7 · Next Mission

Recommended Next Mission

This recommendation comes from CCDR's own recorded next_task, not from Claude directly. You always approve before it becomes a new mission.

—

Timeline

Every entry below is a real, timestamped event — either recorded by your browser as it happened, or read directly from CCDR's stored checkpoint. Nothing here is simulated.

No mission events recorded yet this session.

Previous Missions & Checkpoints

Every checkpoint this project has ever saved, most recent first -- the same resume history CCDR itself reads from.

Stable engineering rules live in these generated files instead of being re-sent in every mission prompt.

Shows what CCDR generated for this mission. Whether each file is present, current or user-authored on disk is verified on your machine at execution time — the control plane cannot read your working tree.

Working context

Reference for development only -- to start a new Claude conversation, use the Bootstrap Prompt from Conversation Reset in Developer Tools instead.

Nothing restored yet.

Built on demand from this project's own checkpoint history -- no filesystem scanning. Kept up to date automatically; this button is only for advanced/manual use.

CLAUDE.md Review

Claude Conversation

Last Completion

No mission has been executed for this project yet.

---

Continue working. No action required.

Estimated from CCDR's own activity in this workspace (prompts sent, time since last checkpoint) -- not a direct read of your Claude conversation.

The project continues -- only the Claude conversation changes. This does not end your session or complete the project.

Conversation Reset Complete

Continue your CURRENT Claude conversation. To start a NEW one instead, use Conversation Reset above.

Attach files

Optimization Profile Standard Balanced Aggressive Caveman This Machine (for the Per-Machine savings breakdown, e.g. "Laptop" or "Desktop")

Paste real terminal/log output here. Cleaning only removes repeated noise -- real failures are always preserved, and the original is always one click away.