chenhg5/cc-connect

What it does

cc-connect is a Go service that sits between an AI coding agent running on your own machine and a messaging platform you already use. Agents such as Claude Code, Codex, Cursor Agent, Gemini CLI, Kimi CLI, OpenCode and any ACP-compatible agent are exposed through chat adapters for Feishu/Lark, DingTalk, Slack, Telegram, Discord, WeCom, LINE, Matrix and others. You send a message, and cc-connect relays it to the local agent CLI and streams the result back into the chat.

The bridge is modular: platform adapters and agent adapters are separate, so one process can host multiple projects, each pairing a different agent with a different platform. Most supported platforms use WebSocket, long polling or Socket Mode, so a public IP is usually unnecessary; LINE webhook and WeCom webhook are the notable exceptions.

Getting started

Order matters: the agent CLI must exist and be authenticated before cc-connect starts, otherwise the service exits with an error such as claudecode: claude CLI not found in PATH.

1. Install at least one agent CLI, for example npm install -g @anthropic-ai/claude-code or npm install -g @openai/codex, then confirm with claude --version.
2. Authenticate it interactively, e.g. claude login.
3. Install cc-connect with npm install -g cc-connect, brew install cc-connect, a binary from GitHub Releases, or build from source with Go 1.22+ via make build.
4. Run cc-connect. It creates ~/.cc-connect/config.toml on first launch and prints a Web admin address such as http://localhost:9820.
5. In the Web UI, create a project, add a platform, paste the bot token from that platform's developer console and save. Config hot-reloads, so you can message the bot and the prompt reaches your local agent.

cc-connect web only opens the configuration UI; the service itself still needs cc-connect running in another terminal. Manual TOML configuration is also supported by copying config.example.toml into ~/.cc-connect/config.toml. Per-platform steps live in the platform setup guides, e.g. Feishu and Telegram.

Operations and controls

Session control happens through slash commands in chat: /new, /list, /switch, /dir to change or reset the working directory, /mode for permission modes including yolo (auto-approve all tools), /model, /provider, /cron for scheduled tasks, and /memory for agent instruction files. /whoami or /status reveals your user ID, which you place in admin_from under [[projects]] to authorize privileged commands such as /dir and /shell.

Projects rotate to a fresh session after inactivity (reset_on_idle_mins, default 30, set 0 to disable) to limit context drift. Generated images and files can be pushed back to the chat with cc-connect send --image or --file, capped at 50 MiB by default and toggleable through attachment_send.

Constraints

• Requires at least one installed and logged-in agent CLI; without authentication the agent rejects every prompt.
• OS-user isolation (run_as_user) works only on Linux/macOS and currently only for Claude Code; it needs passwordless sudo from the supervisor user and a preflight via cc-connect doctor user-isolation.
• Some capabilities are platform-limited: Markdown rendering, images, and voice depend on the vendor API and may need extra [speech] or TTS configuration.
• The repo declares an MIT license in the README, but the GitHub metadata lists no license, so verify the license file before redistribution.
• The README is padded with sponsor blocks and third-party API reseller promotions; these are commercial placements, not part of the documented feature set.

HotRepo's take

cc-connect appeared in a single Show HN thread at launch on 2026-02-28 with 1 point and 2 comments, and HotRepo has no daily-rank or star-snapshot records for it. In that thread the author contrasted the project with a single-platform Discord/Codex bridge, describing the modular adapter approach as the distinguishing choice (reply). The README itself shows a broad support matrix and an active beta changelog, but there is no on-site data here to judge adoption or long-term momentum.