1. WeKan
  2. Platforms
  3. NitroStack

NitroStack · Incubated at WeKan

The MCP platform for secure, reliable production agents.

NitroStack combines an open-source SDK with platform capabilities for building, securing, deploying and operating MCP infrastructure.

01 — What NitroStack enables

The MCP layer between enterprise systems and agents.

NitroStack powers the MCP infrastructure behind our agentic work, giving agents, applications and workflows secure, reusable access to enterprise tools and services.

AI agents and applications

  • AI agents and workflows
  • AI-native applications
  • Copilots and developer tools
  • ChatGPT, Claude, NitroChat and other MCP clients

NitroStackMCP infrastructure and reusable tools

  • MCP servers
  • Tool definitions
  • Security and access control
  • Testing
  • Deployment
  • Reusable integrations

Enterprise systems and services

  • CRM and ERP
  • Databases and data lakes
  • Internal APIs
  • Legacy applications
  • Document systems

MCP (Model Context Protocol) is the open standard AI assistants and agents use to call a product's tools and data.

02 — Inside NitroStack

An integrated experience for enterprise-grade MCP development.

NitroStack brings together the tools and enterprise controls needed to build and operate MCP infrastructure, from identity, access and permissions to testing, managed releases, hosting and observability.

Product experience

  1. Build

    NitroStack SDK

    Build MCP servers and tools in TypeScript or Python.

    • Interactive widgets for MCP apps
    • OAuth 2.1, JWT and API-key auth
    • A CLI to scaffold, run and build
  2. Design and test

    NitroStudio

    Design, test and debug MCP tools and apps in a desktop studio.

    • MCP apps from OpenAPI or Postman
    • Inspector and test cases
    • Skills for multi-step tool workflows
  3. Deploy

    NitroCloud

    Deploy and operate MCP servers without managing infrastructure.

    • Deploys from GitHub, with rollback
    • Autoscaling and custom domains
    • Logs and dashboards
  4. Consume

    NitroChat

    A white-label chat interface for MCP-powered experiences.

    • Your branding
    • Several AI model providers
    • Interactive widgets in the conversation

Enterprise production controlsAcross the platform

  • Identity, access and permissionsOAuth 2.1 with PKCE, JWT and API keys in the SDK, with guards and scopes for permissions.
  • Testing and validationAn MCP inspector and test cases in NitroStudio, and a testing module in the SDK.
  • Managed releases and hostingNitroCloud deploys from GitHub with autoscaling and rollback, or run it in your own cloud.
  • Observability and operationsLog streaming, dashboards and health checks; audit logs on NitroCloud's Enterprise plan.

03 — MCP apps

From MCP tools to interactive applications.

NitroStack helps teams create MCP-powered application experiences, including interactive widgets that agents and compatible clients can present to users.

Illustrative MCP app: interactive UI generated around an enterprise capability.
Expose capabilities
Make enterprise tools and data accessible through MCP.
Generate interactive experiences
Present structured actions and dynamic widgets inside compatible clients.
Use across interfaces
ChatGPT, Claude, NitroChat and custom applications: widgets are compatible with the OpenAI Apps SDK and MCP Apps specifications.

04 — Reusable MCP infrastructure

Build the interface once. Reuse it across agentic experiences.

NitroStack provides a reusable MCP foundation, so new agents, applications and workflows build on established integrations, tools and controls instead of recreating them.

Reused, not rebuilt, by each new agent

  1. 01IntegrationsConnections to enterprise systems, built once and shared by every agent, application and workflow.
  2. 02Tool definitionsThe same tools, names and schemas wherever an MCP client calls them.
  3. 03Authentication and permissionsOne access model that each new agent inherits instead of redefining.
  4. 04Testing and deploymentTests and releases already in place, so each new consumer starts from a production-ready base.

05 — In our work

How NitroStack powers WeKan's agentic engineering.

We use NitroStack across new agentic systems, existing enterprise applications and AI-native products to create reusable, production-ready MCP infrastructure.

New agentic systems

Build the MCP tool layer that connects agents to enterprise systems.

NitroStackA shared tool layer for new agents

Existing agentic systems

Introduce reusable MCP interfaces and standardized access to existing services.

NitroStackStandard interfaces over what exists

AI-native applications

Build products with MCP capabilities available from the first release.

NitroStackMCP from the first release

MCP enablement

Expose existing enterprise capabilities to compatible agents and applications, without a complete agentic systems engagement.

NitroStackYour capabilities, reachable by agents

06 — Extend legacy systems

Extend what can stay. Modernize what can't.

Some legacy systems can be exposed to agents through a secure MCP layer. Others need deeper modernization before agents can rely on their capabilities. WeKan can support both.

Legacy systems

Assess readiness

Path A · Powered by NitroStack

Wrap and extendExpose what can safely remain to agents through MCP, without a full rewrite.

Path B · Nexus

ModernizeTransform what can't support agentic access into a modernized system.

NitroStack MCP layer

Agents and applications

NitroStack adoption

  • 550K+npm downloadsAll-time, for the open-source SDK, CLI and widget packages, to 8 Oct 2026.
  • 2,464GitHub starsAnd 1,424 forks of the open-source repository, 10 Oct 2026.
  • Apache-2.0Open sourceThe license for the SDK, CLI and widgets.

Checked 10 October 2026 on npm and GitHub. nitrostack.ai ↗

07 — Proof

WeKan deliveries on NitroStack.

An insurtech

Insurance quoting, as MCP tools

What was implemented
Quoting, carrier comparison and the route to advisors, exposed as MCP tools.
Outcome
An app inside ChatGPT and Claude API workflows in its .NET platform; customers ask in their own words, with less manual support work.

A Latin American legal-services marketplace

Legal workflows, connected through MCP

What was implemented
Client intake, job records and legal templates, with each workflow step connected through MCP.
Outcome
Two conversational workflows, for client intake and legal documents: more efficient onboarding and simpler document-heavy work.

Build your MCP foundation with WeKan.

We'll send an MCP assessment: the tools we'd build, the auth model and how long it takes. Free, and yours to keep.

Prefer to talk? sales@wekan.ai