v0.6.1

What's new in 0.6.1 (2026-07-12)

These are the changes in xagent v0.6.1.

New features

• feat: add Seedance video generation by @qinxuye in #763
• feat(mcp): connect key-based catalog apps (Google Maps) + shared-env layer by @bluefish-08 in #753
• feat: add embed_origin parameter for iframe authentication by @codeacme17 in #746
• feat: add ElevenLabs TTS support by @qinxuye in #762
• feat(api-keys): support multiple API keys per agent with a centralized management page by @yiboyasss in #748
• feat: admin agent-config visibility and conversation-log rendering by @bluefish-08 in #783
• feat: Implement sandbox management improvements with idle TTL by @codeacme17 in #779
• feat: context-window-based compaction threshold with usage indicators by @bluefish-08 in #786
• feat(core): execution scope context for scoping sandbox by @codeacme17 in #789
• feat(core): decouple sandbox mount root from full workspace segments (scope-prefix mounts) by @codeacme17 in #817
• feat: support ElevenLabs STT by @qinxuye in #793
• feat: add OAuth token resolver hook for MCP configs by @AlexLiu190625 in #814
• feat: add optional quota enforcement hook seams by @bluefish-08 in #841
• feat: add A2A agent protocol support by @qinxuye in #838
• feat(models): add model_id and prompt-cache-hit tokens to usage details by @bluefish-08 in #844
• feat(tools): positive tool-allowlist hook and layered-quota error codes by @codeacme17 in #854
• feat: add ElevenLabs sound effect and music generation by @qinxuye in #853
• feat: serialize task execution control by @qinxuye in #845
• feat: per-step interrupt seam for mid-run quota enforcement by @bluefish-08 in #855

Enhancements

• enh: trigger provisioning, auditing, and performance by @codeacme17 in #765
• enh: dedupe checkpoint trace storage and prune history (20 GB -> 2.2 GB on a real DB) by @qinxuye in #846
• ref(template): change AI Chatbot Agent execution mode by @bsbds in #797
• ref(mcp): single source of truth for catalog app auth classification by @bluefish-08 in #819
• ref(kb): shrink compatibility adapters (H08) by @rogercloud in #839

Bug fixes

• fix(storage): strengthen object-storage key validation by @codeacme17 in #782
• fix(widget): gate App Widget access on a per-agent widget key by @codeacme17 in #781
• fix(migrations): merge alembic heads left by #781 by @codeacme17 in #787
• fix(kb): shrink KBCoordinator to a pure router by @rogercloud in #791
• fix: keep durable file previews independent of storage paths by @qinxuye in #788
• fix: persist connector runtime snapshots for web tasks by @AlexLiu190625 in #784
• fix: scope model providers by category by @qinxuye in #780
• fix(memory): enforce scope on by-id get/update/delete by @codeacme17 in #820
• fix: scan due scheduled triggers in-process so they fire without Celery beat by @bluefish-08 in #813
• fix(core): confine file-id resolution to the active execution scope by @codeacme17 in #829
• fix(sandbox): scoped executions fail closed instead of falling back to local execution by @codeacme17 in #827
• fix(memory): stop dropping data on LanceDB schema mismatch by @codeacme17 in #823
• fix: resolve MCP runtime translation keys by @AlexLiu190625 in #837
• fix(core): bind durable-storage handles to the active execution scope by @codeacme17 in #835
• fix: point homepage docs link to hosted knowledge base by @yiboyasss in #830
• fix(core): re-check containment when resolving relative workspace paths by @codeacme17 in #825
• fix: make task message delivery interrupt-safe by @qinxuye in #840
• fix(memory): push memory-dimension filters into lancedb where clause (#822) by @codeacme17 in #836
• fix: filter TTS voice metadata by configured models by @qinxuye in #848
• fix: preserve OAuth resolver actor diagnostics by @AlexLiu190625 in #851
• fix: deduplicate replayed user messages by @qinxuye in #849

Documentation

• doc(readme): add PyPI badge and pip / curl install methods by @qinxuye in #761

Others

• chore: remove stray superpowers spec and CHANGELOG.md by @qinxuye in #858

Full Changelog: v0.6.0...v0.6.1