LibreChat v0.8.8 released: Agent management API and attached workspaces

LibreChat v0.8.8 introduces an Agent management API (beta), letting developers create, discover, update, and delete agents via OpenAPI and Swagger UI, manage agent files and skills, and authenticate machine clients through deployment-bound OIDC. The new attached workspaces (highly experimental) let each code worker select or save a default workspace so agents can inspect directory trees, read and search files, write changes, and run Bash within bounded timeouts. Background tool controls support canceling ordinary background tools while keeping detached sub-agent execution independent; code approval controls offer ask, allow, and deny options plus a full-access mode. Manual context compaction starts summary-only turns before the context window fills, preserving recent conversation. Context usage can be inspected for conversation, retained tool traffic, agent instructions, cache, cost, and runtime pressure. Unified attachments support single uploads with LibreChat routing content to the model or extracting text, providing file search and code tools as needed. New models GPT-6 Astra and GPT-6.1 Sol support Responses API routing and tool calls. For observability, Trace Viewer inspects ordered model conversations, tool turns, and costs; OpenTelemetry exports correlated app logs, and Langfuse tracing identity and metadata are allowlist-configured. Reliability and security gains include agent continuation and checkpoint recovery, Redis liveness detection, DocumentDB coordination, OpenID and MCP OAuth sessions, shared-link throttling, tenant isolation, attachment boundaries, and upload error handling.