What is an MCP gateway?

An MCP gateway is a single, governed entry point in front of many Model Context Protocol servers. Instead of every agent connecting to every server with its own credentials, the gateway centralizes authentication, observability, and control in one place. Unmeshed acts as an MCP gateway and adds durable execution on top, so tool calls get retried on failure and every invocation stays auditable.

Do my existing MCP servers work with Unmeshed?

Yes. Unmeshed orchestrates and gateways multiple MCP servers, so any server that speaks the Model Context Protocol can sit behind one authenticated, observable front door. From there, you can compose their tools into durable workflows alongside your own.

How is authentication handled?

Auth is centralized at the gateway with enterprise RBAC. Agents authenticate to Unmeshed instead of holding separate credentials for each MCP server, and you manage permissions in one place. Unmeshed runs fully managed in our cloud, or self-hosted in your own infrastructure.

Which MCP clients can call Unmeshed tools?

MCP is an open standard, so any client that implements the Model Context Protocol can discover and call tools exposed through Unmeshed. You can also trigger workflows through REST APIs, webhooks, queues, and schedules, using the Java, Go, Python, or TypeScript SDKs.

How does pricing work?

Pricing is predictable. There are no per-action or per-transition fees, so costs don't balloon as your agents call more tools. There's a free tier to start, no credit card required.

Model Context Protocol · for AI engineering teams

One MCP gateway. Every tool, workflow, and server.

Unmeshed is MCP-native orchestration. Publish durable workflows as MCP tools your agents can call, and put every MCP server behind one authenticated, observable, production-grade gateway.

Start Free Contact Us

From the creators of Netflix Conductor.

refund_flow.py

# refund_flow.py — a durable Unmeshed workflow
client = UnmeshedClient(client_config)

client.register_worker(
    Worker(execution_method=process_refund, name="process-refund")
)
client.start()  # retries, state, and audit handled by the engine

# Published as an MCP tool behind the Unmeshed gateway
# TODO: confirm MCP registration API

# Any MCP client — Claude, or an agent you build — calls it:
# → tools/call
#     { "name": "process-refund",
#       "arguments": { "order_id": "ord_1042" } }
# ← { "status": "COMPLETED", "refund_id": "rf_2210" }
#     durable · retried · fully audited

What is MCP orchestration?

MCP server

A service that exposes tools to AI agents over the Model Context Protocol. Each server publishes what it can do; agents discover and call those tools at runtime.