Show HN: Claudraband – Claude Code for the Power User
TL;DR Highlight
Claudraband is a CLI/library tool that wraps Claude Code TUI, allowing you to maintain sessions and control it headlessly via an HTTP daemon or ACP server. It's worth paying attention to for developers who want to integrate Claude Code into automated workflows.
Who Should Read
Developers who want to leverage Claude Code in continuous session-based automation workflows rather than one-off commands. It's particularly useful for remotely controlling Claude Code in CI/CD or headless environments, or for integrating it with editor plugins.
Core Mechanics
- Claudraband wraps Claude Code's TUI (Terminal UI) with a controllable terminal, allowing you to maintain sessions and resume them later or inject prompts from outside.
- The core features are: resumable non-interactive workflows, remote/headless session control via an HTTP daemon, editor integration via an ACP (Agent Communication Protocol) server, and custom tool integration via a TypeScript library.
- The session resumption feature is key: you can send new prompts while maintaining the context of a previous Claude Code session using the format `cband continue <session-id> 'question'`. This is conceptually an addition of session support to the existing `claude -p`.
- It uses tmux as a backend for managing sessions locally. tmux must be installed to use the first-class workflow.
- It doesn't aim to replace the Claude Code SDK, but is a tool specialized for personal and ad-hoc use. It doesn't touch OAuth or bypass the Claude Code TUI, and always operates after authenticating with Claude Code through an actual Claude Code session.
- Installation is simple with npm/bun, and it bundles `@anthropic-ai/claude-code@2.1.96` within the package. You can override this version with the `CLAUDRABAND_CLAUDE_PATH` environment variable.
- It's currently in an Experimental stage and explicitly states that the project will continue to evolve as Claude Code and ACP clients change.
Evidence
- There was criticism that Claudraband exacerbates Anthropic lock-in due to its reliance solely on Claude Code. Opinions were raised that a more versatile tool supporting Gemini CLI, Codex, and OpenCode is needed.
- While some suggested that Claude Channels could simplify prompt injection and response extraction, limitations were pointed out, such as the need to use the terminal interface for commands like `/clear` due to a poorly designed architecture.
- Concerns were raised in the comments regarding potential conflicts with Anthropic's Terms of Service (ToS) when using subscription-based access. While it explicitly states it doesn't bypass the official TUI, the legality of automated usage remains unclear.
- There was a comment that the license wasn't specified in the README. While a MIT license file exists in the repo, it's not prominently displayed in the README, causing confusion.
How to Apply
- If you repeatedly perform code review or analysis tasks using Claude Code, you can start a session with `cband 'review the staged diff'` and continue subsequent work while maintaining the previous context with `cband continue <session-id> 'keep going'`. This is much more efficient than opening a new session each time.
- If you need to automate Claude Code in a headless server or CI environment, you can use the HTTP daemon mode. The daemon runs like a server, allowing you to control sessions via REST API calls for integration with scripts or other services.
- If you want to integrate Claude Code into your own editor plugin or development tool, you can use the ACP (Agent Communication Protocol) server functionality or directly import the TypeScript library to construct custom workflows in code.
- If you want to quickly test it out, you can run it directly without installation using `npx @halfwhey/claudraband 'review the staged diff'` or `bunx @halfwhey/claudraband '...'`.
Code Example
# One-off execution
npx @halfwhey/claudraband "review the staged diff"
bunx @halfwhey/claudraband "review the staged diff"
# Global installation
npm install -g @halfwhey/claudraband
# Start local session
cband "audit the last commit and tell me what looks risky"
# Check session list
cband sessions
# Resume session (maintain previous context)
cband continue <session-id> "keep going"
cband continue <session-id> --select 2
# Specify custom Claude Code binary path
export CLAUDRABAND_CLAUDE_PATH=/path/to/claude-codeTerminology
Related Papers
FinHarness: An Inline Lifecycle Safety Harness for Finance LLM Agents
금융 AI 에이전트가 실행 중간에 위험한 툴 호출을 차단하면서도 정상 승인율을 유지하는 인라인 안전 프레임워크
Retrying vs Resampling in AI Control
Claude Code처럼 의심 행동을 막고 재시도하는 방식이 오히려 공격자에게 힌트를 줘서 더 위험할 수 있다는 연구.
Mitigating Provenance-Role Collapse in Long-Term Agents via Typed Memory Representation
LLM 에이전트의 장기 메모리가 출처를 뒤섞는 문제를 '타입이 있는 메모리 원자' 구조로 해결한 논문
Push Your Agent: Measuring and Enforcing Quantitative Goal Persistence in Long-Horizon LLM Agents
LLM 에이전트가 '100개 찾아줘'를 실제로 100개 찾을 때까지 멈추지 않게 만드는 방법과 벤치마크.
CoSPlay: Cooperative Self-Play at Test-Time with Self-Generated Code and Unit Test
Ground Truth 없이도 코드와 Unit Test가 서로 평가하며 함께 품질을 높이는 추론 시간 최적화 프레임워크
Multi-Stream LLMs: new paper on parallelizing/separating prompts, thinking, I/O
현재 LLM이 입력·사고·출력을 순차적으로만 처리하는 구조적 한계를 지적하고, 각 역할을 별도의 병렬 스트림으로 분리해 동시에 처리할 수 있는 Multi-Stream 방식을 제안한 논문이다. 에이전트의 효율성·보안·모니터링 가능성을 모두 개선할 수 있다는 점에서 주목받고 있다.