The fragmented landscape of AI coding tools just got a unified front end. Odd Essentials has released Agent Guild, a local web application that orchestrates major coding CLIs including Claude Code, Codex CLI, Gemini CLI, and Grok Build. Rather than forcing users to juggle separate terminal windows, the tool presents each session as a distinct card in a browser interface, allowing developers to monitor active sessions, model usage, and helper agents in real time.

Persistent Terminals and Local Control

A critical technical feature of Agent Guild is its session manager, which owns the underlying pseudo-terminals. This architecture ensures that closing the browser page does not terminate the running processes; sessions continue working in the background and can be resumed later. The tool requires Node.js 22 or newer and supports Windows 10 1809+, macOS 11+, and Linux. It handles the installation of the various CLIs via npm, Homebrew, or WinGet but explicitly avoids handling user sign-ins, keeping credentials separate from the orchestration layer.

Multi-Account Management and Usage Metrics

For developers juggling work and personal subscriptions, Agent Guild supports multiple accounts per tool. Each account maintains its own sign-in, usage meters, and session history. The interface provides granular visibility into rate limits, displaying meters for 5-hour and 7-day windows for supported tools like Claude Code and Codex CLI. It also surfaces plan details, such as extra-usage spend or credit balances, helping users avoid unexpected costs while maintaining high-frequency coding sessions.

Model Benchmarking and Agent Visibility

The platform integrates deep benchmarking data, grading the newest fully benchmarked models on coding, intelligence, and agentic work from S to D. Users can compare models across 13 specific benchmarks, including Artificial Analysis indexes and Design Arena results for web apps and data visualization. Furthermore, the tool visualizes helper agents as 'familiars' on each card, providing transparency into which sub-agents are summoned during a session. This visibility is achieved through hooks that report agent activity directly to the Agent Guild manager.

Key Takeaways

  • Agent Guild unifies Claude Code, Codex CLI, Gemini CLI, and Grok Build into a single local web interface.
  • Sessions are persistent and survive browser closure because a local manager owns the terminals.
  • The tool provides detailed usage meters for rate limits and credit balances across multiple accounts.
  • Model selection is aided by integrated benchmarks and grades across 13 different performance metrics.

The Bottom Line

Agent Guild is a pragmatic solution for the 'CLI fatigue' plaguing modern AI-assisted development. By treating terminals as persistent, observable objects rather than ephemeral windows, it brings much-needed order to the chaotic workflow of switching between competing LLMs.