Show HN: ctx – an Agentic Development Environment (ADE)
TL;DR Highlight
ADE (Agentic Development Environment) is a tool that allows you to run multiple coding agents such as Claude Code, Codex, and Cursor in a containerized, isolated environment from a single interface, and safely merge the results of parallel tasks.
Who Should Read
Development team engineers or platform/security personnel who are simultaneously using multiple AI coding agents (Claude Code, Codex, etc.) within a team, experiencing workflow fragmentation and management difficulties, or wanting to control the file access and network permissions of agents.
Core Mechanics
- ctx positions itself as a tool in the ADE (Agentic Development Environment) category. If IDE provides a workspace for developers, ADE provides a workspace for AI agents. It allows you to integrate and use different coding agents such as Claude Code, OpenAI Codex, and Cursor in a single UI.
- The environment in which the agent runs is isolated in a container, allowing you to explicitly control the scope of disk access and network communication. It acts as a security layer to prevent the agent from unintentionally transmitting data externally or accessing sensitive files.
- Instead of requesting user approval every time an agent works, you can configure it to operate autonomously within a predefined scope of permissions (bounded autonomy), preventing interruptions to the workflow.
- It utilizes Git worktree (a feature that allows you to simultaneously check out a repository into multiple directories) to run multiple agents in parallel on different tasks. Each task proceeds in an independent worktree, enabling concurrent work without file conflicts.
- It features an Agent Merge Queue to sequentially organize the changes from agents working in parallel and merge them into the main branch. Handling conflicts, especially when two agents modify the same file, is a key challenge that ctx aims to address.
- Task-specific execution sessions, diffs, and transcripts (a complete record of the agent's work), along with artifacts, are integrated into a single review screen. This makes it easier to track and review what the agent has done.
- It can be run not only on a local machine but also on remote devboxes or VPSs managed by the user. In a typical local workflow, you can use your own API keys and models without a ctx account.
- Installation is possible with a single curl command (`curl -fsSL ctx.rs/install | sh`). For first-time users, it is recommended to verify the entire loop (provider connection → workspace addition → simple first task execution (e.g., label modification, bug fix)) to validate functionality.
Evidence
- One comment pointed out the limitations in multi-repo environments. In actual development environments, dozens of repositories are often grouped into a single workspace, and it's unclear whether ctx properly supports scenarios where a single feature implementation is planned across multiple repos.
- Several comments pointed out that the GitHub repo (github.com/ctxrs/ctx) lacks source code and only contains links. The repo is marked as 'open source,' but the lack of actual code caused misunderstanding, and questions about the open-source status and free/paid policy followed.
- One comment noted that the Merge Queue, a core feature, is receiving less attention. While parallel agent execution is easy, merging changes when two agents modify the same file without conflicts is a problem that most tools haven't solved. Comments requested specific explanations of the conflict resolution method.
- There were questions comparing it to existing ADE tools like Conductor. Concerns were raised that existing ADEs lack code navigation capabilities and are GUI-focused, and the need for a tool that combines IntelliJ-level IDE functionality with the multitasking capabilities of ADEs was expressed.
- One comment asked if it supports GitHub Copilot subscription accounts (OAuth/HMAC). The practicality of adoption depends on being able to use an existing Copilot subscription because the cost of direct API calls is burdensome.
- A bug report was posted as a comment stating that the app window appears as a blank screen and the menu bar is completely disabled on Linux. Linux support appears to be unstable.
- One user shared an experience of abandoning Conductor because it forced the GitHub PR workflow. They asked if ctx allows free management of local repos and flexible handling of merges.
How to Apply
- If your team uses Claude Code and Codex with fragmented tools, introducing ctx as a common interface can centrally manage task history and diffs in a single review screen, regardless of the agent type.
- If you are concerned about agents freely reading the entire codebase and arbitrarily calling external APIs, you can operate ctx with container isolation and network egress policies to allow agents access only to permitted disk ranges and external communication.
- When you need to develop multiple independent features simultaneously, you can leverage ctx's worktree-based parallel task execution and Agent Merge Queue to create a pipeline where each agent works in a separate branch and merges safely in the order completed.
- When first introducing it, it is recommended to start with tasks that are small in scope and low in risk, such as modifying label text or fixing a simple bug, and then gradually apply it to more complex tasks after verifying the entire loop (install → provider connection → workspace addition → task execution → diff review → finalize).
Code Example
# ctx installation (one-line curl)
curl -fsSL ctx.rs/install | sh
# Subsequent flow (based on documentation)
# 1. Run app and connect provider (Claude Code, Codex, etc.)
# 2. Add local repo or remote devbox/VPS as workspace
# 3. Example of first task execution (recommended to start with low-risk tasks)
# - Modify a label or verification message
# - Fix a narrow-scoped, clear bug
# - Make a small UI/document/configuration change
# 4. Review diff and finalize changesTerminology
Related Papers
Show HN: ctx – Search the coding agent history already on your machine
Claude Code, Cursor, Codex 등 코딩 에이전트가 이전 세션의 논의·결정·실패 시도를 잊지 않도록 SQLite로 인덱싱해 재사용할 수 있게 해주는 오픈소스 CLI 도구다.
Micro-Agent: Beat Frontier Models with Collaboration Inside Model API
vLLM 팀이 단일 모델 API 호출 뒤에서 여러 모델이 협업하는 'Micro-Agent' 개념을 공개했습니다. 별도의 에이전트 코드 없이 라우터 레이어에서 모델 조합을 실행해 GPT-4급 결과를 더 저렴하게 낼 수 있다는 아이디어입니다.
Ornith-1.0: self-improving open-source models for agentic coding
Gemma 4와 Qwen 3.5를 기반으로 파인튜닝한 코딩 특화 오픈소스 모델로, RL(강화학습)을 통해 스캐폴드(에이전트 실행 구조)까지 함께 최적화하는 방식을 주장하지만, 커뮤니티에서는 벤치마크 과최적화에 불과하다는 의심을 받고 있다.
Entity Binding Failures in Tool-Augmented Agents
AI 에이전트가 올바른 도구를 선택해도 잘못된 대상에 실행하는 'Entity Binding 실패' 문제를 정의하고, 이를 막는 실행 정책을 평가한 논문.
Herdr: Agent multiplexer that lives in your terminal
여러 AI 코딩 에이전트(Claude, Codex 등)를 하나의 터미널에서 동시에 실행·관리할 수 있는 Rust 기반 오픈소스 툴로, tmux처럼 세션이 유지되고 SSH로 원격 접속도 가능해 멀티 에이전트 워크플로우를 크게 단순화해준다.
Ornith-1.0: Self-scaffolding LLMs for agentic coding
모델이 문제 풀이 전략(scaffold)을 직접 생성하고 개선하는 자기강화 학습 프레임워크를 적용한 오픈소스 코딩 특화 LLM으로, 9B 소형 모델부터 397B 대형 모델까지 라인업을 갖추고 SWE-Bench 등 주요 벤치마크에서 Claude Opus 4.7을 능가하는 성능을 보여줬다.