likecode

An opinionated Claude Code fork focused on routed models, multi-model coordination, and a denser terminal dashboard.

Origin: likecode is derived from the learn-likecc project.

Current state: custom route models, alias switching, /mmodel orchestration, HUD/rewind overlay, and improved background task visibility are all working in this fork.

What likecode adds

• modelRoutes with short aliases such as mm27, mm25, g5, g51
• /model can switch by alias and shows route source + host
• /mmodel turns natural-language requests into lightweight multi-model orchestration prompts
• route models can carry optional pricing metadata
• startup dashboard shows:
  • built-in models
  • route model alias mapping
  • route source file and host
• double-Esc opens a Rewind | HUD overlay
• footer and background task UI are tuned for multi-agent work
• alt+b opens the tasks panel quickly

Quick Start

Clone the repo:

git clone [email protected]:3873225350/like-code.git
cd like-code

Install dependencies and run in dev mode:

bun install
bun run build
bun run dev

Install cc-route (required for model routing)

like-code delegates model routing to the standalone cc-route project:

# Option A: npm
npm install -g cc-route-proxy

# Option B: pre-built binary (no Node.js)
# See https://github.com/3873225350/cc-route#quick-start

Then configure your routes in ~/.claude/settings.cc-route.json and launch with clauder instead of claude.

See docs/cc-route-integration.md for full details.

For a permissive local dev session:

bun run dev:danger

To build a local standalone binary:

bun run package:local

Run the packaged binary directly:

./dist/likecode

To expose likecode as a global command on your machine:

mkdir -p ~/.local/bin
ln -sf "$(pwd)/dist/likecode" ~/.local/bin/likecode

Make sure ~/.local/bin is on your PATH, then you can start it anywhere with:

likecode

For local debugging with permissive filesystem access:

likecode --dangerously-skip-permissions

Settings Layout And Precedence

likecode currently uses two related configuration layers:

1. the original Claude Code settings layer
2. the likecode-specific extension layer

They are read separately, then applied together.

Standard Claude Code settings

These are the original Claude-style settings files:

• ~/.claude/settings.json
  • user-level global defaults
• .claude/settings.json
  • shared project settings, suitable for committing
• .claude/settings.local.json
  • personal project override, usually gitignored

Effective precedence inside the standard Claude settings layer is:

~/.claude/settings.json
  < .claude/settings.json
  < .claude/settings.local.json
  < CLI --settings
  < managed/policy settings

So for normal Claude settings, local project settings beat shared project settings, and shared project settings beat user-global settings.

likecode settings

These are the likecode-specific files used for routed models and related extensions:

• ~/.claude/settings.likecode.json
  • user-level likecode defaults
• .claude/settings.likecode.local.json
  • project-level likecode override

Effective precedence inside the likecode layer is:

~/.claude/settings.likecode.json
  < .claude/settings.likecode.local.json

So the project's settings.likecode.local.json wins over the user-global settings.likecode.json.

Relationship Between The Two Layers

The important rule is:

• standard Claude settings control the normal Claude Code behavior
• likecode settings add likecode-specific behavior, especially routed-model data

For routed models specifically, the effective order is:

environment variable model routes
  > standard Claude settings modelRoutes
  > likecode settings modelRoutes

And inside the likecode settings portion:

.claude/settings.likecode.local.json
  > ~/.claude/settings.likecode.json

In practice, that means:

• if you want team-shared Claude behavior, use .claude/settings.json
• if you want your own project-only Claude override, use .claude/settings.local.json
• if you want likecode route aliases, custom route hosts, and likecode-specific route metadata for this project, use .claude/settings.likecode.local.json
• if you want your own cross-project likecode defaults, use ~/.claude/settings.likecode.json

Recommended split:

• .claude/settings.json: shared hooks, shared plugins, shared Claude behavior
• .claude/settings.local.json: your personal local Claude overrides
• ~/.claude/settings.likecode.json: your reusable likecode defaults across projects
• .claude/settings.likecode.local.json: the project's routed-model definitions and aliases

Example

Typical project setup:

~/.claude/settings.json
~/.claude/settings.likecode.json
<repo>/.claude/settings.json
<repo>/.claude/settings.local.json
<repo>/.claude/settings.likecode.local.json

Recommended usage:

• keep generic Claude preferences in ~/.claude/settings.json
• keep team-shared project behavior in .claude/settings.json
• keep your machine-specific project tweaks in .claude/settings.local.json
• keep likecode route aliases and route providers in .claude/settings.likecode.local.json

Route Model Config

Put your route models in .claude/settings.likecode.local.json:

{
  "modelRoutes": {
    "MiniMax-M2.7": {
      "alias": "mm27",
      "baseURL": "https://api.minimaxi.com/anthropic",
      "authToken": "YOUR_TOKEN",
      "pricing": {}
    },
    "MiniMax-M2.5": {
      "alias": "mm25",
      "baseURL": "https://api.minimaxi.com/anthropic",
      "authToken": "YOUR_TOKEN",
      "pricing": {}
    },
    "glm-5": {
      "alias": "g5",
      "baseURL": "https://mydamoxing.cn/anthropic",
      "authToken": "YOUR_TOKEN",
      "pricing": {}
    }
  }
}

Then you can use:

/model mm27
/model g5

/mmodel

/mmodel is a prompt-level orchestration helper for multi-model work. It does not hardcode a scheduler; it generates a structured instruction that tells Claude to delegate with explicit model aliases.

Example:

/mmodel 用mm25，以及mm21共同写作完成一个贪吃蛇游戏（源码放在这里templete/v4），mm27监视其完成情况。

Typical behavior:

• mm25 handles implementation
• mm21 handles UI, README, or a second bounded task
• mm27 monitors progress and reports completion

The monitor flow in this fork is tuned to avoid endless waiting by using bounded checks and a report-oriented finish.

Interface Highlights

Welcome screen

• current model + host
• built-in model list
• route model alias mapping
• route source file display
• quick hint for alt+b

Rewind / HUD

Double-press Esc on an empty input to open:

• Rewind: restore or summarize from an earlier point
• HUD: inspect session/project token usage and switch HUD mode

Background tasks

• compact multi-agent footer summaries
• progress bar previews for local agents
• task panel optimized for selecting and drilling into running agents

Scripts

bun run dev
bun run dev:danger
bun run build
bun run package:local
bun run typecheck

Project Notes

This repository started from a Claude Code source study base and is being reshaped into a more experimental operator-focused fork.

Some notable areas in this fork:

• src/utils/model/modelRoutes.ts
• src/utils/model/modelOptions.ts
• src/components/LogoV2/CondensedLogo.tsx
• src/components/MessageSelector.tsx
• src/components/HudPanel.tsx
• src/components/tasks/BackgroundTaskStatus.tsx

Image Assets

README images live in:

docs/images/

Current README screenshots are pulled from docs/images/, and more TUI captures can be added there later.

Disclaimer

This repository includes source material originally studied from a public leak event discussed on 2026-03-31. Original Claude Code source remains the property of Anthropic. This fork is a learning and modification project and is not affiliated with Anthropic.