Embedded auth server: Transparent token refresh and Redis TLS​

The embedded authorization server received two production-readiness improvements, both contributed by Aron at Muon Space:

• Transparent upstream token refresh handles token expiration automatically in the background without interrupting active MCP sessions or requiring users to re-authenticate, so long-running workflows don't break when identity provider tokens expire.
• TLS support for Redis token storage enables encrypted connections between the auth server and Redis, meeting compliance requirements for organizations operating in multi-tenant or zero-trust environments where network-level encryption is mandated by policy.

Cloud UI: Server display names and deep link fix​

Cloud UI now supports display_name for MCP servers, so administrators can assign human-readable titles that appear throughout the interface instead of raw server identifiers. This release also fixes Claude Code deep link generation, which was previously producing incorrect URLs.

Desktop UI: Refreshed navigation​

Desktop UI navbar has been refactored to align with the updated Stacklok brand style, providing a cleaner and more consistent navigation experience across the app.

ToolHive Runtime: Configurable proxy timeout​

The streamable HTTP proxy previously used a hardcoded 30-second response timeout, which caused 504 Gateway Timeout errors for MCP tools that need more time to complete, such as tools querying large datasets or interacting with slow external APIs. The timeout is now configurable via the TOOLHIVE_PROXY_RESPONSE_TIMEOUT environment variable, with a more generous default.

Getting started​

For detailed release notes, check the project repositories:

• ToolHive Runtimes (CLI and Kubernetes Operator)
• ToolHive Desktop UI
• ToolHive Cloud UI
• ToolHive Registry Server

You can find all ToolHive documentation on the Stacklok documentation site.

• On-demand tool discovery means agents no longer receive hundreds of tool descriptions in context. Instead, tools are discovered at request time and only the relevant ones are surfaced (up to 8 by default, configurable) via hybrid (semantic + keyword search), cutting token usage by 60-85% per request while improving tool selection accuracy.
• Team-wide token savings from a single deployment means the 60-85% per-request reduction the Optimizer delivers on desktop now applies to every developer connected to vMCP. Deploy it once, and the savings multiply across the entire team without anyone managing a local optimizer instance.
• No per-developer configuration required. Developers point their MCP client at the vMCP endpoint and get optimized routing automatically, with no local embedding models, no search parameter tuning, and no setup drift across the team.

The setup is GitOps-friendly: EmbeddingServer and VirtualMCPServer CRDs deploy through your existing CI/CD pipeline. For the full configuration reference and quickstart examples, check out the vMCP optimizer guide to get started. For desktop users, MCP Optimizer remains available through the ToolHive UI and CLI as before.

Getting started​

For detailed release notes, check the project repositories:

• ToolHive Runtimes (CLI and Kubernetes Operator)
• ToolHive Desktop UI
• ToolHive Cloud UI
• ToolHive Registry Server

You can find all ToolHive documentation on the Stacklok documentation site.

Registry Server: Multi-namespace management and production hardening​

Registry Server v0.6.0 and v0.6.1 improve operational readiness and simplify production deployments:

• Cluster-wide scanning lets the Registry Server watch MCP servers across multiple Kubernetes namespaces, not just the namespace where it's deployed. Configure which namespaces to monitor with the THV_REGISTRY_WATCH_NAMESPACE environment variable.
• RBAC support in Helm includes the necessary ClusterRole and RoleBinding resources out of the box, so you no longer need to create RBAC permissions manually for multi-namespace deployments.
• PostgreSQL-only storage simplifies configuration by removing in-memory and file storage options; all deployments now use production-grade persistence by default.
• Registry entry separation restructures the internal data model to decouple entries from MCP-specific details, laying groundwork for future extensibility such as non-MCP tool types.

Additional fixes address leader election conflicts when multiple Registry Server instances exist in the same namespace, and add 64 KB metadata size limits to prevent abuse.

Virtual MCP Server: Circuit breakers and observability standards​

vMCP adds resilience primitives and aligns with emerging observability standards:

• Circuit breakers detect unhealthy backends and temporarily remove them from the available capability set. This prevents a single degraded MCP server from causing cascading failures across dependent workflows, particularly important as organizations connect more business-critical tools through MCP.
• OpenTelemetry MCP semantic conventions align ToolHive's telemetry with the official OTel MCP standards merged in January 2026. If you're already using observability tooling like Grafana, Datadog, or Honeycomb, ToolHive traces and metrics will use the same attribute names and span formats the rest of the AI ecosystem is standardizing on.

Embedded authorization server: Federated identity for MCP servers​

The embedded authorization server eliminates the complexity of configuring and managing credentials for authenticated MCP servers. Instead of requiring users to obtain tokens independently and configure credentials locally, ToolHive handles the full OAuth flow:

• Centralized token management: The authorization server runs in-process within the ToolHive proxy, exposing standard OAuth endpoints. Users authenticate through their company identity provider (such as Okta, Entra ID, or Google), and ToolHive issues tokens that MCP servers accept.
• Dynamic Client Registration: MCP clients register automatically without manual configuration at the identity provider, removing the operational burden of pre-registering OAuth clients for each tool.
• Per-request identity: MCP servers receive authenticated identity on each request without users needing to store long-lived credentials on their machines.

This approach supports multiple authentication patterns. For example, with AWS STS token exchange, users authenticate via their company identity provider and receive appropriate IAM roles based on policy, with no AWS CLI setup or stored credentials required.

For enterprise teams, this addresses common security requirements: credential rotation happens automatically, audit trails capture who accessed what, and security policies that prohibit local credential storage are satisfied by design. For details, see the auth framework documentation.

Cloud UI: Faster registry discovery​

Cloud UI v0.2.0 makes it easier to find and evaluate servers in large registries:

• Registry filtering by data source helps you find relevant servers quickly when working with registries containing multiple data sources.
• Tool visibility displays tool names and descriptions for MCP servers directly in the registry view, so you can understand what each server offers before connecting.

Desktop UI: Latest models and simplified maintenance​

The Desktop UI keeps your environment current with less manual effort:

• Opus 4.6 in Playground gives you immediate access to test prompts against the latest Claude model without waiting for broader rollouts.
• Update outdated MCP servers detects locally installed MCP servers with newer versions available and lets you update them directly in the app, eliminating manual version checking and reinstallation.

Ecosystem updates: Broader client compatibility​

The Desktop UI and CLI now support automatic configuration for three additional clients:

• OpenAI Codex
• Gemini CLI
• Mistral Vibe

Register these clients once, and ToolHive automatically manages their MCP server configurations. Whenever you run an MCP server, it's immediately available without manual config file edits. See the client compatibility reference for the full list of supported clients.

Getting started​

For detailed release notes, check the project repositories:

• ToolHive Runtimes (CLI and Kubernetes Operator)
• ToolHive Desktop UI
• ToolHive Cloud UI
• ToolHive Registry Server

You can find all ToolHive documentation on the Stacklok documentation site.

Virtual MCP Server: Dynamic discovery and operational status​

vMCP gains capabilities that eliminate manual intervention and improve visibility:

• Dynamic backend discovery detects servers added or removed from an MCPGroup and updates routing automatically, eliminating pod restarts when changing your MCP backend topology.
• Static configuration mode provides an alternative for security-conscious deployments where vMCP requires zero Kubernetes API access; choose between operational convenience and minimal attack surface based on your security requirements.
• Operational status reporting now surfaces deployment health and backend server status in the CRD status field for improved debugging and monitoring integration.

Registry: OpenTelemetry and tool metadata​

The ToolHive Registry Server expands its observability and developer experience:

• OpenTelemetry support enables integration with your existing observability stack for tracing and monitoring registry operations.
• Tool definitions can now be specified in MCPServer CRDs and are exposed via the API, so MCP server entries in the registry can display their available tools.

CLI and Desktop UI: Header configuration for remote servers​

The CLI and Desktop UI now support custom header configuration for remote MCP connections, useful when services require specific headers beyond standard authentication. Configure non-sensitive headers directly, or reference encrypted secrets for sensitive values like API keys.

Kubernetes Operator: Custom CA support​

The MCPServer CRD now supports custom certificate authorities, enabling ToolHive to validate OAuth tokens issued by private on-premises identity provider instances, essential for organizations with internal PKI infrastructure.

Getting started​

For detailed release notes, check the project repositories:

• ToolHive Runtimes (CLI and Kubernetes Operator)
• ToolHive Desktop UI
• ToolHive Cloud UI
• ToolHive Registry Server

You can find all ToolHive documentation on the Stacklok documentation site.

Virtual MCP Server: Health monitoring and composite tool fixes​

vMCP adds operational capabilities that teams need for production deployments:

• Backend health monitoring detects and handles unreachable servers gracefully, reducing user-facing errors when individual MCP servers restart or become temporarily unavailable.
• Composite tool fixes ensure type coercion in templates works correctly and error continuation behaves as expected, so multi-step workflows complete reliably even when intermediate steps encounter issues.

ToolHive desktop UI: Proxy configuration and stability​

The desktop UI gives you more control over how MCP servers connect:

• Custom proxy ports simplify deployment in environments with specific port requirements or when running multiple instances.
• Stability improvements through upgraded core dependencies (Electron, ai-sdk) and fixes for form validation and environment variable handling reduce crashes and unexpected behavior.

Behind the scenes, we've expanded automated testing with Playwright, and work is underway to align CLI and Desktop UI versioning through bundling.

CLI and desktop UI: New auth option for remote servers​

The CLI and desktop UI now support bearer token authentication for remote MCP servers, enabling secure connections to authenticated endpoints, such as enterprise APIs or hosted services, without manual credential handling in each request.

Cloud UI: Easier client setup​

Cloud UI now displays client-specific configurations for MCP servers (Cursor, VS Code, and Claude Code), making setup easier. For Cursor, a single click opens the configuration directly; for VS Code and Claude Code, copy the ready-to-use config or command.

Getting started​

For detailed release notes, check the project repositories:

• ToolHive Registry Server
• ToolHive Desktop UI
• ToolHive Cloud UI
• ToolHive runtimes (CLI and Kubernetes Operator)
• Virtual MCP Server

You can find all ToolHive documentation on the ToolHive documentation site.

Virtual MCP Server: Audit logging, health checking, and composite tool improvements​

vMCP adds more operational capabilities that teams need for production deployments:

• Health monitoring tracks which MCP servers behind vMCP are available and responding, the foundation for building alerts and dashboards that keep your AI workflows running smoothly.
• Audit logging captures workflow events (started, completed, failed) via CRD configuration, giving you the compliance trail and debugging context required for regulated environments.
• Composite tools are more powerful: Reference deeply nested fields using {{.steps.X.output.field.nested}} syntax, and use the new fromJson function to work with MCP servers (like GitHub) that return JSON as plain text. This enables cleaner conditional logic and more flexible workflows.

ToolHive Registry Server: API management and deployment improvements​

Managing registries is now more flexible and automated:

• API-based management lets you add, update, or remove registries programmatically, useful for CI/CD pipelines and infrastructure-as-code approaches.
• Simplified authentication: Configure auth settings directly in the Registry CRD instead of managing separate configuration files. Database credentials specified in the CRD automatically provision pgpass files, reducing manual setup.
• Kubernetes deployment: A new Helm chart (publishing soon) makes cluster deployments more straightforward.

ToolHive UI: Live registry connections​

The ToolHive UI now connects directly to standards-compliant registry API servers like the ToolHive Registry Server. Instead of importing static JSON files, you get live registry data, making it easier to browse and discover available MCP servers in your environment.

Getting started​

For detailed release notes, check the project repositories:

• ToolHive Registry Server
• ToolHive desktop UI
• ToolHive Cloud UI
• ToolHive runtimes (CLI and Kubernetes Operator)

You can find all ToolHive documentation on the ToolHive documentation site.

ToolHive Registry Server: Smarter discovery and scalable sync​

The Registry Server now features improved auto-discovery capabilities and sync architecture. You can now:

• Automatically detect whether an auto-registered server is an MCPServer, VirtualMCPServer, or MCPRemoteProxy.
• Scale sync jobs across multiple registry instances with an improved sync implementation.
• Deploy without hardcoded namespaces; the registry server now auto-discovers its operating namespace.
• Connect the ToolHive CLI and UI to your Registry Server, bringing effortless discovery and deployment to your custom registry.

Virtual MCP Server: Observability and composite tool improvements​

vMCP gained production-grade observability and extended composite tool capabilities.

You can now:

• Expose OpenTelemetry metrics on the standard Prometheus /metrics endpoint (request counts, latencies, backend health, active connections) and export distributed traces for backend client and workflow calls, enabling comprehensive monitoring, alerting, and debugging.
• Define default results for skippable composite tool steps, ensuring workflows behave predictably across all conditional branches.
• Use non-string arguments in composite tools, enabling structured output references for conditional logic, downstream step arguments, and final outputs.
• Include MCPRemoteProxy resources in MCPGroups, allowing remote MCP servers to aggregate alongside local servers in your VirtualMCPServer configuration.

Getting started​

For detailed release notes, check the project repositories:

• ToolHive Registry Server
• ToolHive desktop UI
• ToolHive Cloud UI
• ToolHive runtimes (CLI and Kubernetes Operator)

You can find all ToolHive documentation on the ToolHive documentation site.

ToolHive Registry Server: Unified registry management and discovery​

ToolHive now includes a dedicated Registry Server that aggregates multiple registry sources, keeps them in sync, and helps teams publish a curated "desired" registry. This enables you to:

• Connect multiple registry backends (Git repositories, local or remote files, internally-managed entries, Kubernetes workloads, or upstream MCP Registry API servers).
• Auto-discover MCP servers deployed in Kubernetes.
• Automatically synchronize changes across registry sources.
• Curate what users see via include/exclude filtering.
• Use an authenticated cloud-based UI to view running MCP servers.

ToolHive UI: Registry groups and tool customization (GA)​

The ToolHive UI gained new features to standardize installs and tailor tool experiences. You can now:

• Install MCP server groups directly from a custom registry.
• Customize tools to disable unnecessary tools and edit their names and descriptions for better AI performance and token/cost savings.

Virtual MCP Server: One endpoint for many servers and workflow automation​

Virtual MCP Server (vMCP) simplifies operating many MCP servers in Kubernetes and building cross-system workflows. It enables you to:

• Aggregate multiple MCP servers behind a single endpoint and automatically resolve tool name conflicts.
• Define declarative, multi-step tool workflows with parallel execution, conditionals, error handling, and approval gates.
• Filter, rename, and wrap upstream tools to enforce policies and present simpler interfaces.
• Pre-configure defaults for common parameters to reduce repetition and prevent mistakes.
• Separate auth boundaries: clients authenticate once to vMCP while vMCP manages backend credentials for SSO, centralized access control, and auditing.

Ecosystem updates​

The ToolHive ecosystem continues to grow with support for new clients and MCP server integrations.

Client support:

• Zed: ToolHive now supports managing MCP configurations in Zed

New MCP servers in the built-in registry:

• PagerDuty: Monitor and manage incidents
• Prometheus: Query metrics and monitor systems
• Redfish (from Nokia): Manage server hardware
• SonarQube: Analyze code quality
• ToolHive Documentation Search: Search ToolHive docs directly

Getting started​

For detailed release notes, check the project repositories:

• ToolHive Registry Server
• ToolHive desktop UI
• ToolHive Cloud UI
• ToolHive runtimes (CLI and Kubernetes Operator)

You can find all ToolHive documentation on the ToolHive documentation site.

Expect regular posts here covering:

• Platform updates and improvements
• Developer resources and guides
• New MCP servers added to the registry
• Community highlights

This is a new initiative, so the format will evolve based on what's useful. If there's something specific you'd like us to cover in these updates, let us know.

Check back next week for the first full update!

— The ToolHive Team