gRPC's strong typing and reflection/descriptor discovery make it a great alternative for the tool calling / MCP. In the first part I'd tried out ListTools + a generic CallTool over gRPC.
Now, I updated and am calling gRPC calls directly (tool → grpc_service/grpc_method) with Protovalidate + CEL for client/server pre-validation.
It helps solve the following issues of MCP : tool poisoning, version updating drift/undocumented changes, weaker trust boundaries, and proxy-unfriendly auth. The recent Vercel mcp-to-ai-sdk and Cloudflare’s Code-Mode are indications that we really want to adopt this kind of strong typing and I think gRPC is a great fit.
A curated list of awesome Model Context Protocol solutions for for building, finding, hosting, learning, securing, and using MCP offerings. Sharing the current list here :
--- Private Registries
Ready-to-use pluggable MCP server implementations where MCP servers and tools are managed by the organization. They usually come with auth, guardrails, observability and more.
Composio (https://composio.dev) - Skills that evolve for your Agents. More than just integrations, 10,000+ tools that can adapt — turning automation into intuition.
Docker MCP Catalog (https://hub.docker.com/mcp) - Ready-to-use container images for MCP servers for simple Docker-based deployment.
Glama (https://glama.ai/) - Managed MCP platform: directories, hosted servers, AI gateway, agents/automations, logging/traceability, and public MCP API.
Gumloop (https://www.gumloop.com/mcp) - Workflow automation platform with built-in MCP server integrations. Connects MCP tools to automate workflows and integrate data across services.
Klavis AI (https://www.klavis.ai/) - Managed MCP servers for common AI tool integrations with built-in auth and monitoring.
Make MCP (https://www.make.com/en) - Integration module for connecting MCP servers to Make.com workflows. Enables workflow automations with MCP servers.
mcp.run (https://mcp.run) - One platform for vertical AI across your organization. Instantly deploy MCP servers in the cloud for rapid prototyping or production use.
Pipedream (https://mcp.pipedream.com/) - AI developer toolkit for integrations: add 2,800+ APIs and 10,000+ tools to your assistant.
SuperMachine (https://supermachine.ai/) - One-click hosted MCP servers with thousands of AI agent tools available instantly. Simple, managed setup and integration.
Zapier MCP (https://zapier.com/mcp) - Connect your AI to any app with Zapier MCP. The fastest way to let your AI assistant interact with thousands of apps.
--- Gateways & Proxies
MCP gateways, proxies, and routing solutions for enterprise architectures. Most also provide security features like OAuth, authn/authz, and guardrails.
Arcade.dev (https://www.arcade.dev) - AI Tool-calling Platform that securely connects AI to MCPs, APIs, data, and more. Build assistants that don't just chat – they get work done.
catie-mcp (https://www.catiemcp.com/) - Context-aware, configurable proxy for routing MCP JSON-RPC requests to appropriate backends based on request content.
FLUJO (https://github.com/mario-andreschak/FLUJO) - MCP hub/inspector with multi-model workflow and chat interface for complex agent workflows using MCP servers and tools.
Lasso MCP Gateway (https://www.lasso.security/) - Protects every interaction with LLMs across your organization — simple, seamless, secure.
MCP Context Forge (https://github.com/IBM/mcp-context-forge) - Feature-rich MCP gateway, proxy, and registry built on FastAPI - unifies discovery, auth, rate-limiting, virtual servers, and observability.
MCP-connect (https://github.com/EvalsOne/MCP-connect) - Proxy/client to let cloud services call local stdio-based MCP servers over HTTP for easy workflow integration.
MetaMCP - Open source. Proxy and aggregate multiple MCP servers into meta-MCPs, and host as SSE/SHTTP/OpenAPI endpoints with middleware, OAuth, and tool management. Stdio MCP servers hosting supported.
Microsoft MCP Gateway (https://github.com/microsoft/mcp-gateway) - Reverse proxy and management layer for MCP servers with scalable, session-aware routing and lifecycle management on Kubernetes.
Traego (https://traego.ai) - Supercharge your AI workflows with a single endpoint.
TrueFoundry (https://www.truefoundry.com/mcp-gateway) - Enterprise-grade MCP gateway with secure access, RBAC, observability, and dynamic policy enforcement.
Unla (https://github.com/AmoyLab/Unla) - Lightweight gateway that turns existing MCP servers and APIs into MCP servers with zero code changes.
--- Build Tools & Frameworks
Frameworks and SDKs for building custom MCP servers and clients
Dummy MCP (https://dummymcp.com/) - Create prototype MCP servers instantly: define tools and mock responses to test LLM interactions and iterate quickly.
FastMCP (https://gofastmcp.com/) - The fast, Pythonic way to build MCP servers and clients with comprehensive tooling.
Golf.dev (https://golf.dev/) - Turn your code into spec-compliant MCP servers with zero boilerplate.
Lean MCP (https://leanmcp.com/) - Lightweight toolkit for quickly building MCP‑compliant servers without heavy dependencies.
MCPJam Inspector (https://github.com/MCPJam/inspector) - "Postman for MCPs" — test and debug MCP servers by sending requests and viewing responses.
mcpadapt (https://grll.github.io/mcpadapt/) - Unlock 650+ MCP tools in your favorite agentic framework. Manages and adapts MCP server tools into the appropriate format for each agent framework.
mcp-use (https://github.com/mcp-use/mcp-use) - Open-source toolkit to connect any LLM to any MCP server and build custom MCP agents with tool access.
Naptha AI (https://auto-mcp.com/) - Turn any agents, tools, or orchestrators into an MCP server in seconds; automates hosting and scaling from source or templates.
Tadata (https://tadata.com/) - Convert your OpenAPI spec into MCP servers so your API is accessible to AI agents.
--- Security & Governance
Security, observability, guardrails, identity, and governance for MCP implementations
Invariant Labs (https://invariantlabs.ai/) - Infrastructure and tooling for secure, reliable AI agents, including hosting, compliance, and security layers.
Ithena MCP Governance SDK (https://www.ithena.one/) - End-to-end observability for MCP tools: monitor requests, responses, errors, and performance without code changes.
Pomerium (https://www.pomerium.com/) - Zero Trust access for every identity - humans, services, and AI agents. Every request secured by policy, not perimeter.
Prefactor (https://prefactor.tech/) - Native MCP Identity Layer for Modern SaaS. Secure, authorize, and audit AI agents — not just users.
SGNL (https://sgnl.ai/) - Policy-based control plane for AI: govern access between agents, MCP servers, and enterprise data using identity and policies.
--- Infrastructure & Deployment
Tools for deploying, scaling, and managing MCP servers in production
Blaxel (https://blaxel.ai/) - Serverless platform for building, deploying, and scaling AI agents with rich observability and GitHub-native workflows.
Dexter MCP (https://www.dextermcp.net/) - Comprehensive directory for Model Context Protocol servers and AI tools. Discover, compare, and implement the best AI technologies for your workflow.
MCP Market (https://mcpmarket.com) - Directory of awesome MCP servers and clients to connect AI agents with your favorite tools.
MCP SO (https://mcp.so) - Connect the world with MCP. Find awesome MCP servers. Build AI agents quickly.
OpenTools (https://opentools.com/registry) - Public registry of AI tools and MCP servers for integration and deployment. Allows discovery and use of AI and MCP-compatible tools through a searchable registry.
PulseMCP (https://www.pulsemcp.com/) - Browse and discover MCP use cases, servers, clients, and news. Keep up-to-date with the MCP ecosystem.
Smithery (https://smithery.ai/) - Gateway to 5000+ ready-made MCP servers with one-click deployment.
While building async custom transport for MCP (Model Context Protocol), I found the official spec for writing custom transports broken, the “concepts” guide overwhelming, and no base interfaces in the python SDK. stdio implementation is trivial, streamable HTTP implementation is huge and nothing in between.
Documented some of the pain points in my journey to write a custom transport layer for MCP.
I'm curious how you're making the polling for results approach work right now? Is it a conditional logic that depends on the result from MCP. Or you let the LLM keep deciding on what to do next?