HotPlex: Unified Gateway for AI Coding Agents

What changed

HotPlex has been released as a self-hosted gateway designed to provide a unified access layer for AI coding agents. This tool aims to make AI coding agents accessible across various platforms and team workflows, including web interfaces, Slack, Feishu, and other enterprise messaging systems. The core functionality revolves around presenting a single, production-ready interface to connect to different AI coding agents, such as Claude Code, Codex CLI, and OpenCode Server, as well as any agent compatible with the Agent Client Protocol (ACP).

Key features include:

• Unified Agent Access: Connect to multiple AI coding agents through a single AEP v1 interface.
• Cross-Platform Integration: Utilize agents from embedded web chats, Slack, Feishu, or Yuanxin, eliminating the need to switch contexts to a terminal.
• Persistent Sessions: Maintain conversations that can survive network interruptions or restarts, with support for streaming long-running tasks and accepting follow-up input.
• Self-Hosted Control: Manage authentication, permissions, audit trails, persistence, metrics, and lifecycle management within a single Go binary.

The latest release, v1.50.2, introduces several improvements:

• True Stop Semantics: The gateway now waits for workers, connections, and event forwarders to be completely silent before confirming a stopped_by_user state, preventing old outputs from contaminating subsequent sessions.
• Cross-Worker Lifecycle Isolation: Stopping, retrying, and shared singleton cleanup for ACP, Claude Code, Codex CLI, and OpenCode Server are now unified through run-level barriers and dispatch gates.
• Actionable Failure Feedback: OpenCode quota or rate-limiting failures now return explicit error codes, and the WebChat provides actionable suggestions for retries and credential checks.

HotPlex is built with security, persistence, and observability in mind, offering features like timing-safe API key checks, permission ceilings, SSRF protection, default SQLite persistence (with PostgreSQL support for shared deployments), structured JSON logs, Prometheus metrics, and OpenTelemetry tracing.

Why it matters for builders

For AI builders and developers, HotPlex offers a standardized way to deploy and manage AI coding agents. Instead of building custom integrations for each AI model or platform, developers can use HotPlex as a central hub. This reduces development overhead and accelerates the adoption of AI coding assistants within teams. The ability to integrate agents into existing communication channels like Slack or Feishu means AI capabilities can be accessed more naturally within daily workflows.

Practical impact

Developers can begin using HotPlex by installing it via a simple curl command for macOS/Linux or a PowerShell script for Windows. The hotplex onboard command guides users through configuration, detecting installed agents and setting up necessary secrets. The hotplex gateway start command launches the service, making it accessible via http://localhost:8888 for the embedded web chat. From there, connections to Slack, Feishu, or Yuanxin can be configured. The project also provides SDKs in Go, TypeScript, Python, and Java for programmatic integration. Teams can explore use cases such as turning AI coding agents into shared team capabilities, enabling remote development access with persistent sessions, or powering automations with natural language requests.

Caveats and source limits

The provided source is a GitHub repository page and its associated README. While it details the functionality, architecture, and installation of HotPlex, it does not include independent benchmark results, pricing information, or specific performance metrics for the latest release. The release date mentioned in the excerpt (August 1, 2026) appears to be a placeholder or future date and does not reflect the current release status of the project. The exact capabilities and limitations of specific agent integrations are detailed in separate documentation referenced within the source, but not fully elaborated here. The source indicates a "fresh release" and lists a latest version as v1.50.2, but does not provide a specific release date for this version.