lookfree/cc-harness

Claims below include README evidence. A matching quote or worker agreement does not establish that a claim is correct. Category majority among 2 workers. Version 1.

A desktop workbench for Claude Code that visualizes subagent topology, token costs, and configuration to improve observability.

Desktop workbench for Claude Code — live subagent topology, token cost breakdown with drill-down, hook sandbox, config audit.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Provides observability into Claude Code's complex multi-session orchestration by visualizing subagent topology, token costs, and configuration layers locally.

CC Harness shows you the topology... It reads your local ~/.claude/ state and makes it legible — judgment and actions stay with you.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Built with Electron, React 18, TypeScript, and React Flow to visualize subagent topology and manage local state.

Desktop: Electron + electron-builder... Frontend: React 18 + TypeScript + Vite... Visualization: React Flow (subagent topology)

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Live subagent call tree rendering a 5-level graph with latency, token cost, and nesting depth.

Live subagent call tree — tails session jsonl in real time and renders the 5-level subagent/workflow graph with per-node latency, token cost, and nesting depth.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Hook sandbox for dry-running hooks with simulated input in an isolated environment.

Hook sandbox — dry-run any hook with simulated input: stdout, stderr, exit code, and the transformed result, without opening a real session.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Token cost breakdown by skills, subagents, MCP, plugins, and base session with drill-down to specific messages.

Token cost breakdown — per-session panel splitting skills / subagents / MCP / plugins / base session. Click a slice to rank that bucket's most expensive turns

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Loop and background task monitor aggregating ScheduleWakeup events with trigger history.

Loop & background task monitor — aggregates ScheduleWakeup events across all sessions, classified as pending / fired / expired, with trigger history for each loop.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Skill trigger analyzer extracting keywords and showing example prompts to activate skills.

Skill trigger analyzer — extracts trigger keywords from each skill (classified by action / technology / format / topic) and shows example prompts that would activate it.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Dependency graph mapping relations between Skills, MCP, Hooks, and Commands into workflow chains.

Dependency graph — maps five relation types (Skills → MCP, Hooks → MCP, Skills ↔ Hooks, Commands → Skills, Commands → MCP)

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Auto-memory diffs showing changes to MEMORY.md before and after dream passes.

Auto-memory diffs — snapshots MEMORY.md before and after each dream pass, showing added / deleted / modified / merged / conflict-resolved changes.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Config layer map displaying Skills, Commands, Agents, and Hooks from user, project, and plugin sources.

Config layer map — Skills / Commands / Agents / Hooks from all three sources (user / project / plugin), with override relationships marked.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Visualizes the live subagent call tree in real time, showing a 5-level graph with latency, token cost, and nesting depth.

Live subagent call tree — tails session jsonl in real time and renders the 5-level subagent/workflow graph with per-node latency, token cost, and nesting depth.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Provides a sandbox to dry-run hooks with simulated inputs, showing outputs and exit codes without executing a real session.

Hook sandbox — dry-run any hook with simulated input: stdout, stderr, exit code, and the transformed result, without opening a real session.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Breaks down token costs by session components and allows drilling down to rank the most expensive turns within specific buckets.

Token cost breakdown — per-session panel splitting skills / subagents / MCP / plugins / base session. Click a slice to rank that bucket's most expensive turns.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Monitors background tasks and scheduled wakeups across sessions, classifying their state and displaying trigger history.

Loop & background task monitor — aggregates ScheduleWakeup events across all sessions, classified as pending / fired / expired, with trigger history for each loop.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Analyzes skill triggers by extracting keywords and classifying them, while displaying example prompts that activate specific skills.

Skill trigger analyzer — extracts trigger keywords from each skill (classified by action / technology / format / topic) and shows example prompts that would activate it.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Maps dependencies between skills, hooks, commands, and MCP servers to assemble numbered workflow chains showing execution flow.

Dependency graph — maps five relation types (Skills → MCP, Hooks → MCP, Skills ↔ Hooks, Commands → Skills, Commands → MCP) and assembles related nodes into numbered workflow chains.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Displays diffs of auto-memory snapshots, showing exactly what was added, deleted, or modified during each memory consolidation pass.

Auto-memory diffs — snapshots MEMORY.md before and after each dream pass, showing added / deleted / modified / merged / conflict-resolved changes.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Visualizes the configuration layer map from user, project, and plugin sources, clearly marking override relationships between them.

Config layer map — Skills / Commands / Agents / Hooks from all three sources (user / project / plugin), with override relationships marked.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Offers a real-time optimizer that reprices Opus tokens at Sonnet rates to show exact potential savings based on actual session data.

Includes a real-time optimizer: repricing the current session's Opus tokens at Sonnet rates, so you see the exact dollar amount you'd save — from real data, not an estimate.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Ensures complete local privacy by only reading local configuration and session files without uploading any data to external servers.

Nothing leaves your machine: CC Harness only reads ~/.claude and your project's .claude directory.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Functions as an observability tool rather than a replacement, leaving the actual conversation interaction and decision-making entirely to the user.

Not a chat client, not a CLI replacement. claudia replaces the CLI's interaction surface; CC Harness does not touch the conversation at all.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Stays aligned with the latest Claude Code features and pricing tiers, calibrated against the official changelog for accurate data representation.

Aligned with Claude Code 2.1.220 (model pricing incl. Opus 5 & fast tier, deprecated-model migration guidance, new sandbox/workflow settings, DirectoryAdded hook, subagent nesting semantics — all calibrated against the official changelog).

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Limitation: Requires Node.js 18+ and the Claude Code CLI installed with a valid ~/.claude/ directory.

Prerequisites: Node.js 18+, Claude Code CLI installed (~/.claude/ exists).

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Limitation: Builds are unsigned, requiring manual steps to bypass security warnings on macOS and Windows.

The builds are unsigned, so the first launch needs one extra step

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Limitation: Requires Node.js 18 or higher and the Claude Code CLI to be installed with a valid configuration directory.

Prerequisites: Node.js 18+, Claude Code CLI installed (~/.claude/ exists).

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Limitation: The web mode variant is limited to read-only browsing and lacks the full features available in the desktop Electron version.

Web mode is read-only browsing.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Limitation: Cannot access or analyze Claude Code sessions that are not stored in the local ~/.claude or project .claude directories.

Nothing leaves your machine: CC Harness only reads ~/.claude and your project's .claude directory.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Limitation: While it visualizes stalled workflows, it does not automatically resolve or fix the underlying issues causing the stall.

When a workflow stalls, you see which branch is stuck.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Limitation: Does not execute commands or modify configurations; it only provides visibility, requiring the user to take manual action.

It reads your local ~/.claude/ state and makes it legible — judgment and actions stay with you.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Alternative to claudia

claudia replaces the CLI's interaction surface; CC Harness does not touch the conversation at all.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

developer-tools