For AI agents: a documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.
Cross-Agent Memory Sharing (Closed Preview) — Agents in the same organization can now build shared context about users. When one agent stores a memory, any other agent in that organization can retrieve it to personalize their own interactions. Memories support user and organization scoping, soft-delete, and are viewable by users via the Human API.
MCP Client Integration — Agents can now use external tools by connecting to remote MCP servers. Tools sync automatically and can be assigned alongside existing REST tools. Pre-configured global connectors (Exa, Tavily) are available out of the box. Supports OAuth, API key, or no-auth authentication with automatic token refresh.
Changed
Agent UUID Now Visible and Copyable — The agent UUID is now displayed in the external agent modal and on the agent details page with a copy-to-clipboard button.
The user profile menu now opens with a subtle animation and chevron indicator, and no longer pre-selects the first item on mouse click.
The search bar and toolbar on list pages (Agents, Tools, Guidelines) now stay fixed at the top while content scrolls independently.
Fixed
Concurrent Request Limits — Under high concurrency, requests could occasionally bypass usage limits. Limits are now enforced atomically.
Security
Guidelines Tenant Isolation — Personal guidelines are now only visible to their owner, and organization guidelines are scoped to the correct organization. Previously, guidelines could be accessible across tenant boundaries.
