Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    MCP Server in Production

    1

    Build, test, secure (OAuth 2.1, multi-tenant) and deploy MCP servers that Claude, ChatGPT and IDEs can depend on, with a server template and a protocol smoke test.

    $12

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    MCP Server in Production

    MCP Server in Production

    Example session with this skill installed

    I need an MCP server in Python that lets Claude look up our product catalogue (search products, get product by SKU) and create a quote draft. It will run on Cloud Run and be used by several customers, each of whom should only see their own price lists. Create the server code in server/ and a short DEPLOY.md.

    • Read your context and instructions
    • Compiled the mcp server in

    server/: registry with acme_search_products and acme_get_product (read_only annotations) and acme_create_quote_draft (local_write); domain errors mapped to ToolError; JSON logs on stderr; Streamable HTTP, stateless; TokenVerifier validates iss/aud/exp/scopes and middleware puts the customer id in a ContextVar that every price-list query filters on (never a tool argument); surface test asserts names, annotations and catalogue size. DEPLOY.md: OAuth bearer validation, Secret Manager references, Cloud Run --timeout 300 --min-instances 1 --concurrency 40, smoke test against the built image, gradual traffic shift.

    test asserts names, annotations and catalogue size. DEPLOY.md

    Connects securely to your tools. The creator never sees your data.

    About this skill

    An MCP server is a public API whose callers are models. It must behave like any production API (versioned, authenticated, observable, tested) and read well to a model (clear names, descriptions, honest errors). This skill covers both.

    What your agent does with it

    • Server structure that scales: one tool registry, MCP-free domain logic, prefixed snake_case wire names decoupled from Python names, catalogue-size test.
    • Annotations done right: read_only(), local_write(), destructive() helpers that follow the spec (destructive/idempotent hints unset on read-only tools).
    • Errors: raise, don't return: a returned {"status": "error"} reaches the client as success. One mapping point turns domain exceptions into actionable isError results; MCPError reserved for protocol failures.
    • Transports: stdio vs Streamable HTTP vs legacy SSE; stateless HTTP behind a load balancer; when you really need sessions and affinity.
    • Auth and tenancy: OAuth 2.1 resource-server pattern, token validation (iss, aud, exp, scopes), subject in a request-scoped context, never as a tool argument; linking third-party accounts; tenant-isolation checklist.
    • Testing: body tests, surface tests, and a protocol smoke test (mcp_smoke.py) that runs initialize → tools/list against any stdio server in any language, validates every tool and catches stdout pollution and signature-hiding decorators.
    • Deployment: Dockerfile, Cloud Run settings that matter for MCP (timeout, concurrency, min instances, gen2 for headless browsers, session affinity), secrets, gradual rollout, observability minimum.

    Included: SKILL.md, server_template.py (MCP Python SDK v2), mcp_smoke.py (standard library), auth and deployment references, and an eval case.

    Part of the Production AI Agents series. Start free with Agent Architecture; pair with Agent Guardrails for approval policy inside the agent.

    Python-first (the smoke test is language-agnostic). Does not choose your tools or write their names, descriptions and schemas in depth (see agent-tool-design). Auth guidance should be checked against the MCP spec revision your clients support.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 1 day ago

    Requires

    Python 3.10+
    Python

    Permissions

    File Scopes

    Your
    project
    folder
    (writes
    server
    code
    tests
    and
    deployment
    notes)

    The smoke test starts the server command you give it locally over stdio. Nothing is deployed unless you run the deployment commands yourself.

    Listed1 day ago

    What's inside

    Frequently Asked Questions

    The official MCP Python SDK v2 (pin mcp>=2,<3). The template notes the two changes needed for SDK v1 (FastMCP).

    Yes. mcp_smoke.py talks JSON-RPC over stdio to any server command and needs only the Python standard library.

    No. Cloud Run is the worked example; the same container and settings map to other platforms.