Send a task from WeChat
Onevium 1.1.23 adds WeChat as an entry point to your Claude Code workspace. Open Channels, create a WeChat bot, scan the QR code, and choose its project and model. Save the connection, then send a short task from WeChat and check that the reply returns to the same private chat.
This connection supports private chats, not WeChat groups or channel bot Teams. The default allowlist includes the account used to scan the code. Text, images, and files are supported; your computer and Onevium must remain running and connected for the assistant to work.
Give each Team member a clear task
Choose Team from the @ menu to organize work into a lead session and real member sessions. Confirm that the team Skill and Sessions tools are available. The lead proposes the work, file ownership, acceptance criteria, and estimated cost before creating members; you can inspect each member's conversation as work proceeds.
Use this for tasks with separate outputs, such as one member reviewing an API and another checking its interface. The cockpit brings member status and reports together, while the lead checks the actual changes before marking work verified. File assignments are coordination rules, not filesystem isolation. Onevium Teams are separate from Claude Code's experimental Agent Teams and channel bot groups.
Call a published assistant through the built-in API
External Access lets a backend send requests to a Onevium assistant. Create an application, add a service with its project, model, and allowed capabilities, then check and publish it. Enable access and use the generated request example to verify the final result. Each application has separate credentials and conversations.
Start locally: built-in access listens on loopback, so the first request can run on the same computer without Docker or a public server. Public HTTPS access and a remote gateway are optional. Results can be queried or observed through SSE; result callbacks require an allowed public HTTPS destination. Receiving a run ID means the request was accepted, not that the task succeeded.
Turn a code review into specific follow-up work
In Review, hover over a code line on the modified side and use the plus button to add a note. The notes button gathers annotations across files, so you can review the feedback before sending it to the assistant. Floating navigation helps move between changes and files.
Review notes use the same composer as ordinary messages. Check any text already in the composer before sending the notes directly: the notes may be sent together with that draft. After the assistant makes a change, inspect the new diff and run the relevant checks.
Update and try one workflow
Find the installers and complete release notes on the official v1.1.23 release page. After updating, start with one small task: a WeChat reply, a two-member Team, a local API request, or a review note. The linked tutorials cover the setup and the result to check for each workflow.