Features & Improvements
- Claude Code overhaul: sessions run under Codexia's own session id and follow
/clear, so the transcript, the live card and the session list stay in sync. - Claude Code: edit a past prompt to branch the conversation; the original session is kept intact and the prompt returns to the composer.
- Claude Code: sub-agents nest under their Task call in the trace, with their prompt, steps and report.
- Claude Code: a composer chip lists MCP servers and plugins that are not working, with the reason and a link to Connectors.
- Claude Code: background agents keep an automation running until they finish, and their turns show in the transcript.
- A shared approval card for Codex, Claude Code and ACP, modeled on the Codex app: the reason as the title, the command under review, Deny (Esc) and a split "Allow once" (Enter) button. Answered requests collapse into a one-line trace row.
- Codex: "Approve for me" runs through Codex auto-review, with an "Approve once" retry after a denial; permission presets match Codex.
- Codex: a thread owned by another client opens read-only, with Retry and Fork chat in the composer.
- Connectors: Keenable search and Cohesivity added; brand icons in the connector registry; the catalog moved to
registry/in MCP Linkerserver.jsonformat.
Design
- One thread column (
max-w-3xl) for messages, composers, permission cards and bot chat, so edges line up across Claude Code, Codex, ACP and bots. - The session, tools and model init card is removed from Claude Code transcripts for a quieter stream.
Fixes & Maintenance
- Claude Code transcripts keep reading a session between turns, so follow-up messages from background agents and permission requests reach the screen; the session list and prompts no longer stall in the browser.