Stacklok

Govern MCP servers and control AI access to enterprise tools and data while keeping deployments in your own environment.

Run and scale governed AI agents across local and Kubernetes environments while retaining flexibility to choose models, tools, and infrastructure.

Run coding agents in isolated microVMs and review their changes before they reach your workspace.

Helps organizations move MCP from planning to production with registry guidance, implementation, workflow support, and ongoing operations.

Centralized curation, deployment, discovery, and lifecycle management of Model Context Protocol servers, giving enterprise teams controlled access to approved tools and data integrations.

Products: Stacklok Enterprise MCP Platform

Controls that secure agent execution and access to tools, data, and models through workload isolation, identity enforcement, access policies, provenance verification, and auditing.

Products: Stacklok Enterprise MCP Platform, Broodbox

Orchestration and composition of domain-specific AI agents, connectors, and automated workflows to create agentic experiences across enterprise functions.

Products: Mecatl

Advisory, implementation, embedded engineering, and ongoing operational support for enterprises deploying MCP registries, integrations, and governed agent workflows.

Products: MCP deployment and enablement services

Stacklok Enterprise MCP Platform

Stacklok Enterprise MCP Platform is a self-hosted platform for enterprises to curate, deploy, and govern Model Context Protocol (MCP) servers. Its registry, Kubernetes-native runtime, gateways, and portal help teams manage internal and remote servers, provide controlled access to tools and data, and integrate with existing identity and observability systems. The platform runs in a customer’s private cloud or Kubernetes environment and supports centralized authentication, authorization, auditing, and policy enforcement. Administrators can curate approved servers and skills, while developers and knowledge workers discover and use resources through supported clients and portal interfaces. Stacklok also describes an LLM gateway for securing agent-to-model actions and applying model-selection policies. The Enterprise MCP Platform is a hardened distribution of ToolHive, Stacklok’s open-source MCP platform; organizations can use the complete platform or selected components.

• Curate Trusted Servers And Skills
  Bring approved internal and external MCP servers and skills into one governed catalog. Administrators can review provenance, configure resources, and control who can discover and use them.
• Manage MCP Servers At Scale
  Deploy and operate MCP servers using Kubernetes infrastructure and management patterns. The operator and runtime help platform teams manage deployments and policies as the server estate grows.
• Contain Server Access And Risk
  Use container isolation and configurable network and filesystem permissions to restrict what each MCP server can access within the deployment environment.
• Centralize Governed Tool Access
  Give users and agents a controlled endpoint to approved MCP servers. Centralized gateway controls support consistent access management across internal and remote servers.
• Enforce Identity-Based Access
  Apply authentication and authorization across MCP access using existing identity providers and policies. Per-user identity and token exchange help downstream systems attribute tool calls to the person who initiated them.
• Keep MCP Workloads In Your Environment
  Deploy the platform within customer-controlled infrastructure to keep MCP workloads and data in the organization’s environment and support its deployment and data-residency requirements.
• Protect Server Credentials
  Store and use API tokens through encrypted storage or 1Password integration instead of placing credentials in plaintext configuration files.
• Trace And Audit Tool Activity
  Send platform telemetry to existing observability systems and maintain centralized records of MCP activity, helping teams review usage and investigate actions.
• Simplify Approved MCP Discovery
  Let administrators prepare approved resources centrally and give users a portal to discover and access available MCP servers, tools, and skills.
• Reduce Irrelevant Tool Context
  Surface tools relevant to the task instead of sending the full tool manifest with every request. This can reduce token usage and limit distractions from irrelevant tool descriptions.
• Connect Supported AI Clients
  Connect supported development clients to governed MCP infrastructure, with automatic configuration for selected clients and manual setup available for other compatible clients.
• Verify Server Origin Before Deployment
  Check MCP server provenance and detect tampering before deployment using integrations with Sigstore and GitHub Attestations.
• Control Model Access And Selection
  Apply policies to agent-to-model actions and define model-selection criteria to support control over model use and associated costs.
• Adopt Components On Your Terms
  Use an end-to-end platform or select components to complement existing infrastructure, building on Stacklok’s ToolHive open-source foundation.
• Run Custom MCP Server Packages
  Run custom servers from container images or supported packages, with automatic container building available for several package types.

Mecatl

Mecatl is an open-source, cloud-native agent harness for running and managing agents, from local development to Kubernetes deployments and fleets. It separates the agent loop, tool calls, and untrusted execution into distinct layers, creating boundaries for isolation, policy enforcement, identity, and auditing. Its provider-independent loop can be assembled from a library of tools and skills, while the same backend supports local and Kubernetes operation.

• Scale Agent Fleets
  Run agents on local infrastructure or Kubernetes, with an architecture designed to support deployments from individual developers through platform teams.
• Isolate Untrusted Execution
  Keep untrusted execution separate from the agent loop and its credentials, creating explicit boundaries for security and platform controls.
• Govern And Audit Agent Actions
  Apply identity and policy controls at the harness boundaries and create audit records through the components that enforce permissions.
• Choose Models And Infrastructure
  Keep the agent loop independent of model providers and infrastructure so teams can change models or vendors without replacing the harness.
• Promote Workflows Across Environments
  Use the same backend locally and on Kubernetes to move workflows from development toward production without changing harness configuration.
• Assemble A Tailored Harness
  Build a harness from a library of tools and skills, with tools and MCP services available through an explicit catalog.
• Connect Multiple Clients
  Let different clients attach to the same agent loop, including terminal, API, Slack, and web interfaces.
• Resume Sessions After Worker Loss
  Persist session state and event history so a replacement worker can resume from the last saved turn if a worker stops.
• Run On Your Infrastructure
  Deploy Mecatl locally or on Kubernetes without relying on Stacklok-managed infrastructure.

Broodbox

Broodbox is an open source project for running coding agents in hardware-isolated microVMs and reviewing every change before it reaches your workspace.

• Isolate Agent Execution
  Run coding agents in hardware-isolated microVMs to keep their execution separate from your workspace.
• Review Changes Before Applying
  Inspect every change made by an agent before it reaches your workspace.

MCP deployment and enablement services

Stacklok helps organizations move MCP from idea to a production-ready registry. Services include advisory support for registry curation strategy and compliance alignment, implementation support for registry setup and policy enforcement automation, and managed services for ongoing maintenance, updates, and monitoring. Stacklok also offers Forward Deployed Engineers who embed with teams to build critical workflows and accelerate adoption.

• Set Up Registry And Policy Automation
  Get help setting up an MCP registry and automating policy enforcement as you move toward production.
• Plan A Governed MCP Registry
  Get advisory support for registry curation strategy and compliance alignment as you plan your MCP deployment.
• Build Critical Workflows With Embedded Engineers
  Work with Stacklok engineers embedded with your team to build critical workflows and accelerate MCP adoption.
• Maintain And Monitor MCP Deployments
  Receive ongoing maintenance, updates, and monitoring to support continued MCP operations.
• Connect MCP Servers To Client Tools
  Use built-in, one-click integration with clients such as Cursor and Claude to simplify access to MCP servers.