OpenClaw 2026.3.24 Brings OpenWebUI Sub-Agent Talk and Native Teams Support.

OpenClaw has shipped version 2026.3.24, a release that deepens its OpenAI API integration and expands native support across enterprise chat platforms. The update enables users to communicate with sub-agents through OpenWebUI, introduces interactive Slack reply buttons, and brings native Microsoft Teams support to the local AI agent runtime.

The headline feature is improved OpenAI API compatibility that lets users talk to sub-agents via OpenWebUI, a popular open-source interface for managing AI models. Sub-agents in OpenClaw act as specialized task executors—one might handle file operations while another manages calendar scheduling. Until now, interacting with these sub-agents required OpenClaw's native interface. The OpenWebUI bridge means users can now orchestrate multi-agent workflows through a familiar UI that many teams already use for model switching and prompt management.

This change fits OpenClaw's broader positioning around model flexibility. The platform runs locally and connects to any LLM backend, which has made it a go-to for teams that want to avoid vendor lock-in or need air-gapped deployments.

Version 2026.3.24 also debuts a dedicated control UI for managing skills and tools. Skills in OpenClaw are Python modules that extend the agent's capabilities—everything from web scraping to database queries. Previously, users edited config files or CLI commands to enable or disable skills. The new UI surfaces active skills, their dependencies, and execution permissions in a single pane.

The timing is notable. As we previously covered, OpenClaw has spent the past several releases tightening security around skill execution after a rough April where rapid plugin changes broke stability. A GUI for permissions and dependencies makes that hardening work more accessible to non-technical users.

Enterprise chat integrations get two upgrades. Microsoft Teams support is now native rather than relying on webhooks or third-party bridges. Teams users can @mention the OpenClaw agent, trigger tasks, and receive threaded updates without leaving the app.

Slack gets interactive reply buttons, a UI pattern that lets the agent present choices as clickable actions instead of typed commands. If OpenClaw suggests three possible calendar slots, users can now tap a button rather than copy-paste a timestamp. It's a small quality-of-life change that reduces friction in high-frequency workflows.

Discord also sees an improvement: automatic thread naming. When OpenClaw spawns a thread to handle a multi-step task, it now assigns a descriptive name based on context—"Deploy API to staging" instead of "Thread 1." It's another incremental UX win that matters more in active servers where dozens of threads pile up.

OpenClaw continues to emphasize its architecture: a single local runtime that speaks to any chat platform and any model. Version 2026.3.24 extends that promise into more corners of the enterprise stack. For teams juggling Slack, Teams, and Discord while switching between GPT-4, Claude, and local Llama models, the value prop is consolidation without compromise.

The release is available on GitHub. Users running earlier versions can upgrade through the built-in updater or pull the latest image if running containerized deployments.