Frequently Asked Questions

Find answers to common questions about MCPLambda.

General

What is the Model Context Protocol (MCP)?

The Model Context Protocol (MCP) is an open standard that enables AI models (like Claude, GPT-4, etc.) to securely interact with external tools, APIs, and data sources. Think of it as a universal driver for AI agents.

Why should I use MCPLambda instead of running my own server?

Deploying MCP servers manually involves managing infrastructure, setting up secure networking, handling stateful connections, and configuring secrets. MCPLambda automates all of this, providing a “Vercel-like” experience for backend agent infrastructure — plus a registry, CLI, tool analytics, and an MCP control plane for agents.

Deployment & State

What is the difference between stateful and stateless deployments?

  • Stateless: The server doesn’t remember anything between requests. This is ideal for simple data fetching or time-conversion servers.
  • Stateful: The server needs to maintain data across restarts (e.g., a local database or a session history). On MCPLambda, you can request persistent storage for stateful servers with a single click.

How do I connect my agent to my newly deployed MCP server?

Copy the unique Deployment URL from your dashboard (or mcpl deploy output) and plug it into your MCP client (Cursor, VS Code, Claude Desktop, etc.). MCPLambda serves Streamable HTTP or SSE depending on your transport configuration. Most modern clients support these natively. See Connecting AI Clients.

What deployment auth types are available?

  • oauth (default) — OAuth 2.1 protected endpoint. Best for interactive clients; the client handles the login flow and you usually do not need a deployment API key.
  • key — API key auth. Set a duration (e.g. 30days). Clients send Authorization: Bearer <key>. Retrieve the key from the dashboard or CLI after deploy.
  • none — no authentication (use carefully; typically only for demos or private networks).

CLI example: mcpl deploy … --auth-type key --api-key-duration 30days.

Registry

What is the MCP Server Registry?

A curated Official and Community catalog of MCP servers (ToolHive-backed). Browse, search, and filter in the dashboard, via public API, CLI (mcpl registry), or the MCPLambda MCP tools search_registry / get_registry_server. See MCP Server Registry.

Do I have to use the registry?

No. Bring-your-own-server is first-class: deploy any supported package URL, Git repository, or container image. The registry is discovery and faster install — the catalog is growing and will not list every server you might need.

Is the registry API authenticated?

Reads are public. GET /v1/registry/servers and GET /v1/registry/servers/{name} require no auth. Creating a deployment from a registry entry still uses your normal authenticated deploy path.

CLI

How do I install and use the mcpl CLI?

# macOS / Linux
curl -fsSL https://mcplambda.io/mcpl/install.sh | sh

mcpl login
mcpl deploy npx://@mcp/server-time

Windows: irm https://mcplambda.io/mcpl/install.ps1 | iex. Full reference: The mcpl CLI.

Does the CLI support the registry?

Yes: mcpl registry search <query> and mcpl registry info <name>. Deploy with mcpl deploy using the package or image from the entry.

Analytics

Where do I find tool-usage analytics?

Open a deployment → Analytics tab. You get call volume, error rates, p50/p95 tool-execution latency, and a searchable invocation log. Details: Tool-Usage Analytics.

Why doesn’t analytics show LLM latency or cost?

v1 measures tool-execution latency at the ToolHive proxy (how long the MCP server took). It does not see the model or client E2E path, and does not report tokens or dollar cost. Data is near-real-time (seconds of lag), not a live websocket, and is scoped to one deployment at a time.

MCPLambda MCP Server

What is the MCPLambda MCP server?

A hosted management MCP endpoint at https://api.mcplambda.io/mcp. Connect with a service account token (mcpl_sat_… from Settings → API Tokens) so your AI client can list/create deployments, manage secrets, and search the registry. Guide: MCPLambda MCP Server.

How is a service account token different from a deployment API key?

  • Service account token (mcpl_sat_) — authenticates to the MCPLambda platform (control plane): deploy, secrets, registry tools, etc.
  • Deployment API key (auth_type: key only) — authenticates MCP clients to your deployed server. Not used for platform APIs.

Pricing & Resources

What are “Compute Units”?

Compute Units are a flexible resource pool. Instead of buying a fixed number of servers, you get units (e.g. 10 on Team) and spend them on Small (1 unit), Medium (2 units), or Large (4 units) profiles. See Server Profiles & Pricing.

Can I upgrade my plan later?

Yes, you can upgrade or downgrade your plan at any time. Changes are prorated.

What happens if I go over my unit limit?

We will notify you when you approach your limit. If you exceed it, we provide flexible overage pricing based on the size of the extra server profiles you deploy.

Security

What runs my MCP server under the hood?

MCPLambda runs every deployment on ToolHive, an open-source runtime purpose-built for MCP servers. ToolHive wraps each server in an isolated container with minimal permissions, network access filtering, and an SSE proxy so container ports are never exposed directly to the network.

How are my API keys stored?

Your API keys are stored as Secrets in our encrypted database. They are decrypted on demand only when needed in the dashboard UI and are securely injected into your running deployment as environment variables — ToolHive handles the final step of making them available inside the sandboxed container without surfacing them to the host.

Is my code private?

Yes. MCPLambda provides isolated environments for every deployment. For GitOps flow, we use temporary build environments that are wiped after the container image is created. Every running server lives in its own ToolHive-managed container, so workloads cannot observe or reach each other.

Still have questions? Reach out to our team at hello@mcplambda.io or join our community Discord.