OpenClaw releases 2026.3.24 with Teams, OpenWebUI, and skill-control UI

The core release is about making the OpenClaw runtime easier to drop into existing chat surfaces. OpenClaw's launch post highlights five user-facing changes: OpenWebUI can now "talk to sub-agents," Slack gets interactive reply buttons, Teams is now native, Discord threads can be named automatically, and the control UI adds skill and tool management.

The most implementation-relevant change is the OpenAI-compatible API layer. In the release notes, OpenClaw says it now supports /v1/models and /v1/embeddings, and forwards explicit model overrides through both /v1/chat/completions and /v1/responses. That matters for client compatibility because many off-the-shelf chat front ends, RAG systems, and agent frameworks assume those endpoints exist and that model selection survives proxying.

The admin story here is narrower but practical. The release notes say the control UI now shows only the tools available to the active agent by default, with an optional detailed mode and an "Available Right Now" section. OpenClaw also says bundled skills can now be installed with one-click recipes through the CLI and control UI when dependencies are missing.

On the enterprise chat side, the biggest concrete upgrade is Teams. According to the release notes, OpenClaw moved to the official Teams SDK and added streaming 1:1 replies, welcome cards with prompt starters, feedback hooks, status updates, typing indicators, native AI labeling, and support for message editing and deletion. That expands where agents can run inside organizations, but it also creates more integration points to govern across Teams, Slack, Discord, and OpenWebUI.

A separate summary thread describes this release as more than a connector update, listing ClawHub-first installs, native skill search and update flows, a new plugin SDK, Matrix support, unified image generation, Exa/Tavily/Firecrawl search, OpenShell, and SSH sandboxes for "safely run code." Those claims fit the direction of the official notes, but most are only named in the thread, so the implementation details are still thin.

The practical use case is already visible in the community. In a CRM walkthrough, one user shows an OpenClaw agent wired to Gmail, WhatsApp, Calendar, and a Google Sheet to "auto-update leads" and "auto-send follow-ups," effectively replacing a paid CRM with a chat-driven agent flow. That kind of workflow becomes easier to deploy when the runtime can present itself through OpenAI-style endpoints on one side and native workplace chat clients on the other.