Repository navigation
Major new features
- ⏰ Automations are more flexible and powerful: choose which MCP services each run can use, share automations with others as read-only, and find those "Shared with me" tab. Trigger polling can use a cheaper model than the sessions it starts, and instructions can now be up to 16,000 characters (#12865, #12847, #14065, #13750, #13991, @aravind-segu, @daniellok-db, @omni-resolve-agent[bot])
- 🔲 Session grids are on by default, so you can drag sidebar sessions into a conversation to watch up to four live chats side by side. Workspace content search is also on by default with Cmd/Ctrl+Shift+F (#13845, #13846, @hzub)
- ⏪ Use
/rewindin Claude SDK and Claude Code native chats to go back to an earlier prompt, edit it and resend it. You can also reference workspace files with@in the Claude SDK composer (#12402, #13351, @jsonmp-k8, @ytelang) - 🔍 Better PR review: send local inline review comments to your agent as chat context, and view only the changes since one of a PR's commits to review stacked PRs on their own (#12319, #13206, @daniellok-db, @omni-resolve-agent[bot])
Breaking changes
- Agents on native harnesses no longer get Omnigent's
sys_os_*tools and use the harness's own shell and file tools instead (#14306, @fanzeyi) - Hardened policy checks for Cursor SDK sessions on Linux and macOS. Set a Cursor API key and keep the default
os_envsandbox; agents that declare terminals aren't supported on this harness. (#14308, @fanzeyi)
Bug fixes
- 🔋
omnigent hostno longer polls tmux for idle terminals, which cuts CPU and battery drain. A host whose credentials are rejected now shows offline and asks you to runomnigent login, and multi-line environment values no longer breakomni host enable(#13056, #12563, #14315, @TomeHirata, @omni-resolve-agent[bot]) - 🚢 Docker and Databricks App deployments, and
omnigent serverwithout-c, now apply the full server config. Image and PDF previews and file downloads no longer return HTTP 500 on Starlette 1.8+ (#11105, #12840, #12362, @andrewreid, @natsuskaizen, @omni-resolve-agent[bot]) - 🤖 Interrupting a Codex, Claude SDK or Pi agent no longer drops its harness session. Codex sessions also survive brief host stalls and keep receiving replies when Codex refuses to list earlier turns (#12483, #13789, #13718, @omni-resolve-agent[bot], @aravind-segu, @TomeHirata)
- ⏱️ Agent-created automations keep their
max_cost_usdcap, continue-mode automations recover after a sandbox that never connected, and trigger worker sessions use the automation's permission mode (#8951, #13575, #13717, @omni-resolve-agent[bot], @raj-suket) - 🔁 On multi-replica servers, messages to a sleeping managed sandbox reach the replica that owns its host, in-flight MCP tools keep their results across runner reconnects, and Switch host stops the old runner first (#13795, #13860, #12181, @Edwinhe03, @TomeHirata, @vicennial)
💜 Thanks to our community
This release was shaped by the people who filed issues, opened PRs, and talked through feature requests with us on our Discord! Thank you for building omnigent with us, keep the bug reports, ideas and contributions coming :)
@ajayalfred, @andrewreid, @aravind-segu, @bbqiu, @btli, @Calvinjmin, @craiuconstantintiberiu, @daniellok-db, @dbczumar, @dependabot[bot], @dhruv0811, @Edwinhe03, @fanzeyi, @felho, @hzub, @jbdevprimary, @jsonmp-k8, @Larmstrong1127, @lavneethora, @Leemoonsoo, @LouisLotter, @markhaines, @mh-avant, @michaelp-db, @mxatone, @natsuskaizen, @nihit, @olafurpg, @omni-resolve-agent[bot], @omnigent-ci[bot], @pandalaohe, @prakhar10, @raj-suket, @randypitcherii, @s-sanjay, @samhavens, @samiyerunkar, @serena-ruan, @TomeHirata, @vasvlasov, @vb-dbrks, @vicennial, @yaoharry, @ytelang, @zhengwin, @zozo123, @ZygisKrus
Full Changelog: https://github.com/omnigent-ai/omnigent/blob/main/CHANGELOG.md