Today, we're open-sourcing Volcano SDK, a TypeScript SDK for building AI agents that combines LLM reasoning with real-world actions through MCP tools.
Why Volcano SDK? One reason: because 9 lines of code are faster to write and easier to manage than 100+.
Without Volcano SDK? You'd need 100+ lines handling tool schemas, context management, provider switching, error handling, and HTTP clients.Â
With Volcano SDK: 9 lines.
Look how we compress 100+ lines with the following example:
import { agent, llmOpenAI, llmAnthropic, mcp } from "volcano-ai";
// Setup: two LLMs, two MCP servers
const planner = llmOpenAI({ model: "gpt-5-mini", apiKey: process.env.OPENAI_API_KEY! });
const executor = llmAnthropic({ model: "claude-4.5-sonnet", apiKey: process.env.ANTHROPIC_API_KEY! });
const database = mcp("https://api.company.com/database/mcp");
const slack = mcp("https://api.company.com/slack/mcp");
// One workflow
await agent({ llm: planner })
.then({
prompt: "Analyze last week's sales data",
mcps: [database] // Auto-discovers and calls the right tools
})
.then({
llm: executor, // Switch to Claude
prompt: "Write an executive summary"
})
.then({
prompt: "Post the summary to #executives",
mcps: [slack]
})
.run();
That's it. GPT-5 queries the database, Claude writes the summary, and it posts to Slack. Context flows automatically between steps. Tool schemas are discovered automatically. Errors are retried. All in 9 lines of actual workflow code.
✅ 4 lines of setup — Two LLMs, two MCP servers
✅ Multi-LLM workflow — GPT-5 analyzes, Claude writes
✅ Automatic tool selection — SDK picks the right tools
✅ Chainable steps — Read like a story
✅ Context flows automatically — Each step sees previous results
✅ Production-ready — Built-in retries, timeouts, telemetry
Start building AI agents at volcano.dev.
At Kong, we build a lot of AI agents. Each time, we ended up writing the same code: managing conversation history, handling tool schemas, switching providers, implementing retries, connecting to MCP servers.
We looked at existing options:
We just wanted to chain LLM calls with different providers and use MCP tools without writing infrastructure code every time. It should be simple:
So, we built Volcano SDK. And now you can start building with it.
There are a few core concepts that you will find across Volcano that separate it from the pack:
We wanted to write agents like we write promises: one elegant chain that does exactly what it says, with the right tool for each job:
wait agent({ llm: planner })
.then({ prompt: "Analyze data", mcps: [db] })
.then({ llm: executor, prompt: "Write summary" })
.then({ prompt: "Post to Slack", mcps: [slack] })
.run();
Volcano is focused on one thing: multi-step workflows that mix different LLMs and MCP tools. It's TypeScript-first, type-safe, and small enough to understand in an afternoon, and it comes packed with easy-to-use features and multi-LLM support.
It’s time to start delivering on the agentic promise. It’s time to start building powerful AI agents with the Volcano SDK at volcano.dev. We can’t wait to see what you build.
As of September 2025, Kong Gateway Enterprise 3.8 will enter its End Of Life (EOL) phase and will no longer be fully supported by Kong. Following this, Kong Gateway Enterprise 3.8 will enter a 12-month sunset support period, focused on helping cus
We're very excited to announce Kong Mesh 2.12 to the world! Kong Mesh 2.12 delivers two very important features: SPIFFE / SPIRE support, which provides enterprise-class workload identity and trust models for your mesh, as well as a consistent Kuma R
Ever feel like you're fighting an uphill battle with your API strategy? You're building APIs faster than ever, but somehow everything feels harder. Wasn’t API-first  supposed to make all this easier? Well, you're not alone. And now industry analys
It’s been almost a year since we released our Konnect Terraform provider . In that time we’ve seen over 300,000 installs, have 1.7 times as many resources available, and have expanded the provider to include data sources to enable federated managem
Kong-quer the Agentic AI Hackathon 🚀 Calling all builders, tinkerers, and API innovators. The Kong Hackathon is back for API Summit 2025 ! This year, we’re challenging developers worldwide to create projects that don’t just react, they think , a
In the last two parts of this series, we discussed How to Strengthen a ReAct AI Agent with Kong AI Gateway and How to Build a Single-LLM AI Agent with Kong AI Gateway and LangGraph . In this third and final part, we're going to evolve the AI Agen
In my previous post, we discussed how we can implement a basic AI Agent with Kong AI Gateway. In part two of this series, we're going to review LangGraph fundamentals, rewrite the AI Agent and explore how Kong AI Gateway can be used to protect an LLM