Source: github.com/openclaw/openclaw/releases/tag/v2026.9.36
The project describes this release as its current equivalent of a long-term support line. It carries OpenClaw changes from the end of September 2026, along with critical security updates, reliability and performance fixes, and support for new models.
Failed updates are now easier to recover from. A failed npm rollback no longer leaves the installed package missing, and update checks keep Doctor recovery guidance. Workspace restrictions under tools.fs.workspaceOnly now apply to tools built through Gateway loopback and skill dispatch, which require approval, and media tools stay inside the workspace.
Several parsers now cap how much work untrusted input can trigger. These cover Matrix media metadata, IMAP mail, the TOML key parser, URI parsing, oversized collections in custom A2UI schemas and Discord's embedded SDK. The release also lets xAI models accept off for older models without configurable reasoning effort.
Why it matters if you build agents
Teams using Teams allowlists or workspace restrictions with OpenClaw should recheck their configuration, since unresolved entries now grant nothing and Gateway-built tools now need approval as intended.
Because the release is positioned as the stable line, teams that want fewer changes between upgrades can pin to it and still receive security and reliability fixes.
Get stories like this every weekday
One short email with the day's agent tooling news.