# Zero Cold Starts: Serverless AI Agents on Akamai Functions — Thorsten Hans **URL:** \ **Category:** Content Creators\ **Tags:** functions, mcp, agents, ai-engineer\ **Created:** [10 October 2026 18:00 UTC](https://www.artofsm.art/t/zero-cold-starts-serverless-ai-agents-on-akamai-functions-thorsten-hans/25437 "2026-10-10T18:00:38Z")\ **Posts on this page:** 2\ **Page:** 1
**Author:** ![artesia](https://www.artofsm.art/user_avatar/www.artofsm.art/artesia/32/36_2.png) [@artesia](https://www.artofsm.art/u/artesia)\ **Post date:** [10 October 2026 18:00 UTC](https://www.artofsm.art/t/zero-cold-starts-serverless-ai-agents-on-akamai-functions-thorsten-hans/25437/1 "2026-10-10T18:00:38Z")
[![](https://www.artofsm.art/uploads/default/original/3X/8/5/85825837045059f5aa36cbec3165d80be82002c5.jpeg "Zero Cold Starts: Serverless AI Agents on Akamai Functions — Thorsten Hans") ](https://www.youtube.com/watch?v=Zf_e45GiNCE) Thorsten Hans presents an edge-native architecture using Spin and Akamai Functions to build ultra-fast, globally distributed serverless AI agents and MCPs with zero cold starts, leveraging WebAssembly for portability and low latency. This approach simplifies development and deployment by combining Spin’s vendor-agnostic framework with Akamai’s managed edge platform, enabling scalable, secure, and efficient AI applications accessible worldwide. ---
**Author:** ![artesia](https://www.artofsm.art/user_avatar/www.artofsm.art/artesia/32/36_2.png) [@artesia](https://www.artofsm.art/u/artesia)\ **Post date:** [10 October 2026 18:22 UTC](https://www.artofsm.art/t/zero-cold-starts-serverless-ai-agents-on-akamai-functions-thorsten-hans/25437/3 "2026-10-10T18:22:44Z")
Thorsten Hans, a senior developer advocate at Akamai, presents a solution for building ultra-fast, globally distributed AI agents and MCPs (Multi-Component Processes) using Spin and Akamai Functions. Traditional regional architectures often suffer from high latency and cold start delays, which degrade user experience and increase operational costs. To overcome these challenges, Thorsten introduces an edge-native architecture leveraging WebAssembly (Wasm), which offers small, portable, and fast-executing binary files that can run efficiently across distributed environments. Spin, a CNCF project, is a vendor-agnostic framework that enables developers to build WebAssembly applications using any language that compiles to Wasm, such as Rust, TypeScript, or Python. Spin provides a CLI, a Wasmtime-based runtime, and language-specific SDKs to streamline development. Akamai Functions complements Spin by offering a fully managed, globally distributed serverless platform with zero cold starts, enabling applications to launch in under half a millisecond. It also includes a built-in multi-tenant key-value store for persistent data storage, simplifying deployment and scaling without additional configuration. Thorsten demonstrates how to build an MCP server using WasmCP, a CLI tool that integrates with Spin to handle non-functional requirements like OAuth 2.0 authentication, allowing developers to focus solely on business logic. He walks through creating a simple MCP server locally, testing it, and then deploying it globally on Akamai Functions. The deployment process involves compiling the application to Wasm, uploading it to Akamai’s edge network, and distributing it across service regions worldwide, making the MCP server accessible with low latency to users everywhere. The talk also covers building AI agents using Spin and popular SDKs like the Vercel AI SDK. Thorsten showcases a simple game agent that interacts with a large language model hosted on a GPU-enabled Linode instance. By deploying this agent on Akamai Functions, the system achieves fast response times (around 800 milliseconds) despite the complexity of calling an external LLM. The approach highlights the benefits of compiling interpreted languages into Wasm to optimize cold start and execution performance, leveraging an optimized JavaScript runtime called Starling Monkey. In conclusion, Thorsten emphasizes the portability, speed, and developer experience benefits of using Spin and Akamai Functions for serverless AI agents and MCPs. The open standards-based approach reduces time to market, eliminates infrastructure management, and ensures global low-latency access. Developers can build, test locally, and deploy with simple commands, while Akamai’s globally distributed platform handles scaling and security. For further resources and community engagement, Thorsten directs viewers to the Akamai Developer Hub and invites questions at the Akamai booth. ## Useful Links - [Spin WebAssembly Framework](https://spinframework.dev/) — Central to building and running MCP servers and AI agents as described in the video. - [Spin GitHub Repository](https://github.com/spinframework/spin) — Provides source code and documentation for Spin, directly discussed in the video. - [Akamai Functions Product Page](https://www.akamai.com/products/akamai-functions) — Platform used to deploy and run the WebAssembly MCP servers and AI agents globally as demonstrated. - [wasmcp GitHub Repository](https://github.com/wasmcp/wasmcp) — Key tool demonstrated for building MCP servers focusing on business logic, central to the talk. - [Vercel AI SDK](https://ai-sdk.dev/) — Used to build AI agents deployed on Akamai Functions, part of the demonstration.