hello2cc

To ensure compatibility with various Claude Code versions and naming conventions, hello2cc implements robust identity and context resolution:

Subagent Identity Resolution

The system is designed to be resilient to changes in key naming (e.g., camelCase vs snake_case). It recognizes the following identifiers for subagents:

• agentId
• agent_id
• agent.id

Transcript Path Resolution

Instead of relying solely on inferring paths from subagents/agent-<id>.jsonl, the system supports direct usage of agent_transcript_path or agent_transcript_path found within transcript records.

Context Window Inference

hello2cc uses inferContextWindowSize() to approximate the context window based on model names, mimicking Claude Code's internal logic. It supports common patterns like:

• [1m]
• opus-4-6
• sonnet-4-6
• claude-sonnet-4
• It also respects the CLAUDE_CODE_MAX_CONTEXT_TOKENS environment variable.

The ccstatusline-bridge.mjs script performs the following logic to ensure accurate status reporting:

1. Intercepts JSON: Reads the JSON payload passed from Claude Code to StatusLine.
2. Aggregates Usage: Uses transcript_path to sum up usage from both the main session and any referenced subagent transcripts.
3. Field Mapping: Recognizes various naming conventions for agents and transcripts, including:
   • agentId / agent_id / agent.id
   • agentTranscriptPath / agent_transcript_path
4. Conditional Backfilling: Only fills in context_window fields if they are missing or equal to 0. It will not overwrite non-zero statistics provided natively by Claude Code.
5. Context Window Inference: If context_window_size is missing, it prioritizes the CLAUDE_CODE_MAX_CONTEXT_TOKENS environment variable. If that is also missing, it infers the window size based on common 1M token model defaults.
6. Passthrough: Passes the enriched JSON to ccstatusline@latest without modifying the original source code of Claude Code or ccstatusline.

It is critical to understand that hello2cc is not a full implementation of the hellox Rust workspace. It functions exclusively as a plugin adaptation layer for Claude Code.

What hello2cc handles:

• Hook Lifecycle: Managing events like SessionStart, UserPromptSubmit, SubagentStart, TaskCreated, etc.
• Parameter Normalization: Unifying Agent, Team, and Task semantics.
• Session/Transcript State: Parsing transcripts and extracting host-state snapshots.
• StatusLine Context: Bridging ccstatusline usage by backfilling data from transcripts when Claude Code payloads are incomplete.
• Subagent Context: Injecting context for subagents and handling identity resolution.

What hello2cc does NOT handle:

• Rust Crates: It does not implement hellox-cli, hellox-gateway, or hellox-tui.
• Gateway/Providers: Third-party adapter layers are outside this repository's scope.
• Cloud Services: It does not implement hosted authentication, remote hubs, web consoles, or managed session synchronization (it follows a local-first philosophy).
• Direct Source Modification: It does not modify the Claude Code or ccstatusline source code directly.

When using hello2cc with third-party models, Claude Code's StatusLine may occasionally report usage metrics (like context_window.current_usage, used_percentage, total_input_tokens, and total_output_tokens) as 0. Because ccstatusline prioritizes these fields, the progress bar may incorrectly show 0 even if real usage data exists in the transcript.

Since hello2cc cannot automatically override the existing statusLine.command via plugins, you must manually update your Claude Code settings to use the hello2cc bridge command. This bridge intercepts the JSON sent to StatusLine, aggregates usage from the main session and subagent transcripts via transcript_path, and fills in missing or zeroed usage data before passing it to ccstatusline.

{
  "statusLine": {
    "type": "command",
    "command": "node "D:/GitHub/dev/hello2cc/scripts/ccstatusline-bridge.mjs""
  }
}

To use hello2cc to make third-party models (GPT, DeepSeek, Gemini, etc.) behave more like native Claude Opus sessions within Claude Code, follow these steps:

1. Prerequisites:

   • Node.js 18 or later.
   • Claude Code with plugin support.
   • A working third-party model mapping layer (e.g., CCSwitch).

2. Installation:

   • Clone the repository and enter the directory:

     git clone https://github.com/hellowind777/hello2cc.git
     cd hello2cc

   • Add the local marketplace entry (replace <repo-path> with your actual local path):

     claude plugins marketplace add "<repo-path>"

   • Install and enable the plugin:

     claude plugins install hello2cc@hello2cc-local
     claude plugins enable hello2cc@hello2cc-local

   • Reload Claude Code to apply changes:

     /reload-plugins

3. Verification: Run claude plugins list to ensure hello2cc@hello2cc-local is installed and enabled.

git clone https://github.com/hellowind777/hello2cc.git
cd hello2cc
claude plugins marketplace add "$(pwd)"
claude plugins install hello2cc@hello2cc-local
claude plugins enable hello2cc@hello2cc-local
/reload-plugins

The hello2cc Native style is designed for users who want to maintain a native workflow within Claude Code. It prioritizes native tools, native agent/planning behaviors, and table-friendly outputs. It is an execution-first style that avoids unnecessary ceremony, self-referential narration, and over-guidance.

Key Principles

• Execution-First: Keep interactions restrained and focus on results. Avoid "I'm thinking" or "Let's..." preambles.
• Scope Adherence: Stay within the requested scope. Do not refactor unrelated code or invent future-facing abstractions.
• Directness: Answer questions directly. Avoid front-loading replies with long procedural narration. Answer first, then add necessary details.
• No Invitational Filler: Do not end responses with phrases like "if you want, I can..." or "let me know if...". Instead, end with the result, current status, a blocker, or a concrete next action.
• Faithful Reporting: Report outcomes accurately. If a validation step wasn't run or a check failed, state it plainly.

Formatting Preferences

• Tables: For comparisons, trade-offs, or capability boundaries, use the pattern: one-sentence judgment + compact Markdown table + recommendation.
• Markdown Tables: Prefer standard Markdown tables over ASCII tables or diagrams unless Markdown is insufficient or plain text is explicitly requested.
• Conciseness: Use concise, natural, and accurate wording. Avoid jargon-heavy phrasing, sales tones, or empty intensifiers.

hello2cc is a plugin for Claude Code designed to align the behavior of third-party models (like GPT, DeepSeek, Gemini, etc.) with Claude Code's native Opus session patterns. It improves tool selection, agent routing, and output style.

Prerequisites

• Node.js 18 or higher
• A version of Claude Code that supports plugins
• A model mapping layer (like CCSwitch) if you are not using native Claude models.

Installation Steps

1. Clone the repository:

   git clone https://github.com/hellowind777/hello2cc.git
   cd hello2cc

2. Add the local marketplace: Replace <repo-path> with the absolute path to your cloned hello2cc directory.

   claude plugins marketplace add "<repo-path>"

3. Install and enable the plugin:

   claude plugins install hello2cc@hello2cc-local
   claude plugins enable hello2cc@hello2cc-local

4. Reload plugins:

   /reload-plugins

Verification

Run the following command to ensure the plugin is active:

claude plugins list

Expected result: hello2cc@hello2cc-local should be listed with a status of enabled.