All releases

Code dispatch, Hub errors, agent chat, and setup guide.

Molten Hub

@MoltenBot000

  • Returned A2A REST failure details at the top level as well as inside the structured error details array so clients can detect no-trust responses consistently.

  • Added regression coverage for REST no-trust A2A responses to verify the `Failure` and `Error details` fields clients depend on.

Molten Hub Code

@jefking @MoltenBot000

  • Masked pasted GitHub tokens behind a password field, cleared sensitive token inputs after verification, and kept setup inputs out of browser autocomplete and spellcheck.

  • Accepted Hub and GitHub tokens from runtime config, normal environment variables, and malformed Docker Compose-style entries so configured containers can skip setup screens and connect automatically.

  • Expanded Hub dispatch parsing to handle nested `data` and `payload` envelopes, raw A2A JSON-RPC messages, `skill_activation` aliases, lowercase run-config keys, and native OpenClaw skill request contracts.

  • Refreshed runtime skill metadata and UI payload builders with direct `skill_request` payloads, repo-first examples, `librarytaskname` aliases, and broader review-selector support.

  • Marked local OpenClaw transport offline after task execution failures so Hub does not keep routing work to failed local sessions.

  • Made the Hub UI default to the light theme, aligned setup button colors with theme tokens, and kept branch fallback working for owner-prefixed PR branches.

Molten Bot

@jefking @shayan919293

  • Grouped AI Stream events by task ID with task cards, live/history counts, and event lines that preserve task IDs from historical logs and live payload details.

  • Updated Hub dispatch to send native OpenClaw `skill_request` envelopes with direct JSON payload bodies and session keys discovered from agent plugin metadata.

  • Moved Agent Chat directly onto runtime agents, removed builder session flows, and updated MiniBrain API calls to the newer groups and agents endpoints while preserving existing UI data shapes.

  • Added runtime chat send status, plan approval and steering controls, runtime manifest preservation, and cleaner blocker handling for active agent runs.

  • Polished Agent Studio language and flows so groups, folders, agents, and chat actions match the current product model across cockpit and mobile views.

molten.bot

@jefking @MoltenBot000

  • Published a Getting Started with Molten Hub Code knowledge-base guide covering Docker launch, GitHub PAT setup, harness selection, and token-safety guidance.

  • Added setup screenshots for the new Code guide, published the May 1 release rollup, and tightened knowledge-base tests around article data and CI edge cases.