v0.114.4
Agenta-AI/agentav0.114.4Sep 1, 2026by mmabrouk
AI Summary
This release focuses on improving agent recoverability, enhancing UI discoverability, and fixing critical sandbox and credential management issues. It also raises the free-credit limit and removes experimental components.
Key Highlights
- Increased free-credits per-key TPM cap to 1,000,000
- Improved sandbox reliability by fixing credential substitution errors and detecting stuck sandboxes
- Enhanced UI with discoverable keyboard shortcuts and a split Tools section
- Improved agent recoverability by making the platform readable for AI agents
New Features
- Make agenta.ai readable and recoverable for AI agents
- Make the playground's keyboard shortcuts discoverable
- Stop evicting warm sessions when an integration is added
- Detect a stuck-substitution sandbox before the first model call
- Split the agent Tools section into Integrations and Subagents
- Add a 404 page across OSS, EE, and mobile
Full Release Notes
## What's Changed * [feat] Make agenta.ai readable and recoverable for AI agents by @ashrafchowdury in https://github.com/Agenta-AI/agenta/pull/6361 * [feat] Make the playground's keyboard shortcuts discoverable by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/6373 * [fix] Raise the free-credits per-key tpm cap to 1,000,000 by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/6363 * [fix] Stop blaming the user's key for unsubstituted sandbox credentials by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/6362 * [test] Gate blocks when a model switch loses the warm sandbox by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/6371 * [fix] Key the lifecycle capability table by the wire harness spellings by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/6364 * [chore] Remove the pi_agenta harness experiment by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/6367 * [feat] Stop evicting warm sessions when an integration is added by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/6368 * [feat] Name what evicted a warm session, and when Secrets moved by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/6369 * [feat] Detect a stuck-substitution sandbox before the first model call and rebuild it by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/6370 * [fix] A live-route repair no longer clears every other eviction check by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/6372 * [fix] A cross-modality model switch keeps the warm sandbox by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/6374 * [fix] A changed agent artifact now evicts the warm session by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/6398 * [fix] One harness normalizer, shared by every identity view by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/6399 * [fix] Close two smaller lifecycle gaps: per-turn endpoint eviction, apply-live facet by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/6400 * [feat] Split the agent Tools section into Integrations and Subagents by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/6375 * [feat] Add a 404 page across oss, ee and mobile by @ashrafchowdury in https://github.com/Agenta-AI/agenta/pull/6360 * [docs] Move GitButler stack mechanics out of AGENTS.md into a skill by @ashrafchowdury in https://github.com/Agenta-AI/agenta/pull/6359 * [docs] Update API reference docs by @github-actions[bot] in https://github.com/Agenta-AI/agenta/pull/6357 * chore: v0.114.4 follow-ups (client regen, final error copy) by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/6401 * [test] Four standing gate checks from the free-credits incident by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/6402 * [fix] Keep the run credential when a localhost API URL is bridge-rewritten (F1) by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/6407 * [fix] Classify a direct-provider credential race honestly, once per session (F6) by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/6408 * [fix] Refuse a malformed harness kind with a structured error at commit and invoke (F4) by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/6409 * chore(gate): silence a synthetic placeholder fixture in the secret scan by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/6410 * [fix] Sanitize subagent tool names to the provider pattern (E4) by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/6412 * chore(api): move InvalidAgentHarnessError to the workflow type module by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/6421 * [fix] Address the CodeRabbit review on the credential-race and subagent-name fixes by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/6422 * [chore] Address the CodeRabbit review on the standing gate checks by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/6423 * [release] v0.114.4 by @github-actions[bot] in https://github.com/Agenta-AI/agenta/pull/6366 **Full Changelog**: https://github.com/Agenta-AI/agenta/compare/v0.114.3...v0.114.4