Changelog

What’s new in Muxtra.

Product milestones, safety improvements, and meaningful changes to the workflow.

v0.1.0-beta.1 · Task titles

Short titles, intentional prompts

Muxtra now keeps task tracking separate from the detailed request you give an agent.

  • Positional titles supplied to muxtra start and muxtra team now label workspaces without becoming agent prompts.
  • Added the explicit --prompt option for users who want to send an immediate first request.
  • Added muxtra update for updating npm installations from the beta release channel.
v0.1.0-beta.0 · Public beta

Different models for design and code

Muxtra is focusing its supported release on the CLI and adding responsibility-aware parallel work without replacing the providers’ native interfaces.

  • Added machine-local design and code lane defaults with independent providers and exact model selection.
  • Added muxtra team to create linked UI/frontend and backend/logic tasks in separate workspaces.
  • Added lane-specific agent context and model-aware Claude Code and Codex launches.
  • Focused the supported release and documentation on the CLI workflow.
Composition

Verified work can come back together

The core workflow moved beyond isolated task creation into guarded local composition.

  • Added muxtra finish to verify clean, committed, fresh work and record its exact commit.
  • Added muxtra combine and combine-status with a disposable integration workspace.
  • Combined checks now run before the primary project advances; failures leave it untouched.
Muxtra

A simpler task-first workflow

The project became Muxtra and shifted the everyday interface toward plain-language tasks.

  • Added muxtra setup and muxtra start as the task-first path for creating managed branches and worktrees.
  • Added integration-readiness checks and clearer human-facing task status.
  • Kept the lower-level worktree commands available for agents and advanced workflows.
Coordination

Claims make overlapping intent visible

Isolation prevents overwrites, but it cannot reveal two agents changing the same conceptual boundary. The claim registry fills that gap.

  • Added read and write claims with heartbeats and expiration.
  • Added muxtra who for live ownership visibility.
  • Added claim-aware build diagnostics that relate failing file paths to active work.
  • Kept claims advisory so coordination never becomes an unrecoverable lock.
Foundation

Agent-aware launch context

Muxtra began supplying coding agents with the project workflow and the isolated workspace they should operate in.

  • Added provider-aware launchers and portable instructions.
  • Added repository-owned project configuration.
  • Added managed development server processes, logs, ports, and healthchecks.