The Extending IDE
for & Codex
Make Claude Code intuitive. An agent orchestration tool that unifies sessions, memory, agents, and skills.
A Complete Agent IDE
Multi-Session Management
Run multiple concurrent sessions per agent with real-time status monitoring. Easily switch between sessions and track history.
Agent Memory
Automatically extract and accumulate knowledge from conversations. Remember user info, feedback, and project context.
Agent Management
Define specialized agents in Markdown and compose them as sub-agents. Set model, tools, permissions, and memory per agent to build autonomous AI teams.
Skills System
Define recurring workflows as skills. Bundle instructions, scripts, and templates into reusable packages — invoked manually or triggered automatically.
Scheduled Execution
Auto-run agents on interval, daily, or weekly schedules. Enable unattended operation for recurring tasks.
Org Chart View
Visualize parent-child relationships between agents. Create, edit, and remove sub-agent compositions directly from the org chart.
Integrated Terminal & Editor
Built-in terminal with theme support and a syntax-highlighted editor. Everything Claude Code, in one app.
Manage agents, not files
What's next
Early Access
- Multi-session & resume
- Agent memory
- Agent & skill management
- Org chart view
- Scheduled execution
- Integrated terminal & editor
Next
- Codex CLI support
- Agent & skill sharing hub
- Session analytics & cost tracking
Future
- Windows / Linux support
- Cloud sync
- Team workspaces
Frequently asked
Yes, fully compliant. Conclawd runs the official Claude Code binary as a child process via PTY. It never extracts or reuses OAuth tokens — all API calls are entirely delegated to Claude Code itself. Technically, it's identical to running the claude command in Terminal.app. Conclawd's only role is to provide a GUI layer over the CLI's input and output.
Yes, Conclawd is unaffected. In January 2026, Anthropic blocked third-party tools like OpenClaw from using OAuth tokens server-side, updated its terms of service in February, and formally banned third-party tool usage with Claude Pro/Max subscriptions in April. OpenClaw extracted Claude Code's OAuth tokens and called the API directly via its own HTTP client — it's now detected and blocked by Anthropic's client fingerprinting. Conclawd takes a fundamentally different approach: it runs the official Claude Code CLI as-is via PTY as a child process. It never extracts or reuses OAuth tokens, and never intervenes in API requests or authentication. Technically, it's identical to running the claude command in Terminal.app — which is not prohibited.
We believe the risk is extremely low. Anthropic's ban targets tools that extract and reuse OAuth tokens in unofficial clients. Conclawd simply runs the official claude CLI binary as a child process — it never extracts, stores, or repurposes any tokens. This is technically equivalent to running Claude Code from Terminal.app or iTerm2. Unless terminal emulators themselves are banned, there is no reason for Conclawd to be blocked. It also automatically stays compatible with Claude Code updates, so there are no compatibility concerns.
Free during Early Access. Claude API usage costs follow your existing Anthropic account (Pro / Max / API key).
No. Conclawd runs Claude Code directly. If you're already logged in via Claude Code, no additional auth setup is needed.
All data — including agent configs, memory, session history, and your input — is stored exclusively on your local machine. Currently, no data of any kind is sent to Conclawd servers. In the future, we may introduce optional crash reporting and bug diagnostics, but only with your explicit consent.
macOS only for now. Requires macOS 14+ and Apple Silicon. Windows / Linux support is under consideration.
Currently Claude Code only, but Codex CLI support is planned. The CLI provider abstraction is already designed — agent definitions will stay in a shared format, and you will choose which CLI to run at session launch.
Yes. Conclawd acts as a GUI wrapper for the Claude Code CLI, so Claude Code must be installed beforehand. If Claude Code is not detected on first launch, you will be guided through the installation process.
Yes. Conclawd reads directly from your project's .claude/agents/ and .claude/skills/ directories. Any agents or skills you've already created with Claude Code work in Conclawd with no additional setup.
Yes. Conclawd is open source on GitHub (github.com/btopia1128/conclawd-ide). Bug reports and feature requests are welcome via GitHub Issues.
A new way to use
AI Agent
Download Conclawd for free and start using it today.