v0.106.0

Agenta-AI/agentav0.106.0Jul 28, 2026by bekossy

AI Summary

Focuses on session history reliability, multi-gate approvals, and durable cross-device agent sessions.

Key Highlights

  • Reliable multi-gate approvals with persisted answers
  • Durable cross-device agent session list (archive, ended, titles)
  • Approval friction UX and context-driven config sections
  • Last-message-only turns with runner history rebuilding

New Features

  • Multi-gate approvals
  • Durable cross-device sessions
  • Approval UX improvements
  • Session history reliability

Full Release Notes

## What's Changed
* feat(runner): reliable multi-gate approvals: truthful results, persisted answers, per-card dispatch by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/5382
* feat(frontend): approval friction UX and context-driven config sections (Arda's port) by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/5438
* fix(website): restore GA4 alongside PostHog (analytics continuity) by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/5467
* docs(agent-workflows): reject-siblings approval policy proposal (Deny/Stop/steer) by @ardaerzin in https://github.com/Agenta-AI/agenta/pull/5444
* [chore] Zero out web tsc errors and fail builds on type regressions by @ardaerzin in https://github.com/Agenta-AI/agenta/pull/5464
* fix(frontend): show loading state on home "Create agent" button by @ashrafchowdury in https://github.com/Agenta-AI/agenta/pull/5475
* fix(website): structural accessibility (landmarks, icon-link names, dropdown state) by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/5469
* fix(frontend): union day-of-month and day-of-week in cron next-runs preview by @ardaerzin in https://github.com/Agenta-AI/agenta/pull/4798
* fix(website): open off-site links in a new tab by @ashrafchowdury in https://github.com/Agenta-AI/agenta/pull/5474
* fix(hosting): wire the session history flags into compose and the env examples by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/5488
* [feat] Last-message-only turns: runner rebuilds history from records by @ardaerzin in https://github.com/Agenta-AI/agenta/pull/5486
* fix(sessions): paused-turn transcript hygiene + pause-sentinel replay nudges by @ardaerzin in https://github.com/Agenta-AI/agenta/pull/5451
* fix(website): correct footer social icons + polish HowItWorks scroll by @ashrafchowdury in https://github.com/Agenta-AI/agenta/pull/5473
* fix(runner): keep the warm session when the client sends a minimal history by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/5490
* fix(runner): stop replaying the current prompt as a prior turn by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/5489
* fix(api): attach SessionsConfig so smart truncation stops dropping records by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/5491
* fix(api): order session records by producer time so turns stop interleaving by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/5494
* style(agent-chat): format transcriptToMessages.test.ts with prettier by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/5495
* [feat] Durable cross-device agent session list (archive, ended, titles) by @ardaerzin in https://github.com/Agenta-AI/agenta/pull/5479
* feat(agent-chat): warm Stop + approval dock polish (steer UI flag-gated) by @ardaerzin in https://github.com/Agenta-AI/agenta/pull/5477
* feat(agent-chat): batch approvals — Deny all + context peek [blocked on runner collect window] by @ardaerzin in https://github.com/Agenta-AI/agenta/pull/5470
* fix(runner): fail the turn instead of answering with a lost conversation by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/5493
* feat(sessions): sessions integration — storage rework, reliable approvals, and approval UX by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/5436


**Full Changelog**: https://github.com/Agenta-AI/agenta/compare/v0.105.9...v0.106.0