Wukong Code

Resume the task. Loop until the Gate passes.

Wukong Code is a terminal-native AI coding agent built around one workflow: Resume → Loop → Gate. Continue unfinished work, run the repository's real checks, review the result from a fresh context, and iterate until the outcome is actionable.

• Terminal-native: launch an interactive TUI or run one-off prompts without leaving your shell.
• Loop until verified: wukong loop writes, runs real project checks, scans risks, asks a fresh-context read-only reviewer, and fixes what remains.
• Cross-Agent Resume: continue local Codex, Claude Code, or Cursor sessions from the Wukong TUI without modifying the source session.
• Actionable result: every Loop ends as PASS, NEEDS_WORK, or ERROR, with structured blockers and stable process exit codes.
• Bring your own provider key: Wukong uses the model API you configure; login is used only for the Free Loop allowance and account state.

---

Install

macOS / Linux

curl -fsSL https://wukong.today/install.sh | sh

Or download the latest binary for your platform from the releases page.

Windows

Download the latest .exe from the releases page and place it on your PATH.

Verify

wukong --version

---

Quick start

# Start the interactive TUI
wukong

# In the TUI, continue a local session from another coding agent
/resume codex
/resume claude
/resume cursor

# Run a one-off prompt
wukong -p "explain this repository"

# Iterate until the independent delivery gate passes
wukong loop "finish the current change"

# Verify current changes before committing
wukong verify

# Scan for delivery risks
wukong scan

# Check guard status
wukong guard --status

---

Current commands

Command	Description
wukong	Launch the interactive TUI
wukong -p <prompt>	Run a single prompt and print the result
wukong provider	Manage LLM providers
wukong login	Authenticate via device-code flow
wukong verify	Advanced local deterministic checks used by Loop
wukong scan	Advanced local risk scan used by Loop
wukong guard	Inspect or run the command risk guard
wukong proof	Advanced local verify + scan summary used by Loop
wukong judge	Evaluator-only merge decision
wukong loop <objective>	Run the Loop-first delivery workflow; supports --max-iterations, --every, --model, and --role
wukong today	Show the Daily Proof Briefing and manage today's focus
wukong server	Start the local REST/WebSocket server
wukong web	Open the local web UI
wukong doctor	Validate configuration files
wukong export [sessionId]	Export a session ZIP
wukong vis [sessionId]	Open the session visualizer
wukong migrate	Migrate legacy Wukong data
wukong upgrade / wukong update	Upgrade to the latest version

Inside the TUI, /resume, /resume codex, /resume claude, /resume cursor, and /loop are the primary workflow. Advanced local diagnostics remain available through /verify, /scan, /proof, /judge, and /guard.

Version 0.0.14 provides one two-iteration Guest trial. Signed-in Free users get 10 Loop sessions per month with up to five iterations each. Internal verify, scan, and proof checks do not consume extra quota. Loop exits with 0 for PASS, 1 for NEEDS_WORK, 2 for ERROR, 3 for auth/quota rejection, and 130 when interrupted.

Run wukong --help for the full option list.

---

Documentation

Full docs: docs.wukong.today

• Getting started
• Command reference
• Configuration
• Updates and announcements

---

Support

• Website: wukong.today
• Issues: github.com/mutnpc/wukong-code/issues
• Email: support@wukong.today

---

License

This is proprietary software. See LICENSE.md.