A node-based workflow tool as an agent runtime
Using n8n as the "tool provider" for an LLM agent. Sub-workflows become tools the model can call. Changing agent capabilities becomes a workflow edit, not a deploy.
ReadHyderabad · Full-stack & AI-agent platforms
Siva Prasad Raju Bandi.
Senior full-stack engineer with thirteen-plus years across backend services, web apps, and AI-agent platforms. This is where I write about architectures I've shipped, patterns I've come to trust, and side-projects I return to on weekends.
Recent writing
All posts →Notes from the workbench
Two MCP server patterns: OAuth passthrough vs identity-aware search
Building Model Context Protocol servers for two very different jobs — fronting a SaaS with an OAuth broker, and exposing an internal knowledge base to authenticated users.
ReadDurable log plus event bus: a notification service for AI agents
One service becomes the audit trail and the dispatcher. How persistence-before-fanout, HMAC replay protection, and presigned-URL attachment sync fit together.
ReadBuilding an agent orchestrator without a framework
A hand-rolled state machine for LLM-driven chat: intent routing, tool invocation, SSE streaming, and out-of-band completion — and why we skipped LangGraph.
ReadDesigning an intent classification service for an LLM assistant
The front-of-funnel decision — direct answer, clarify, or call a tool — and why building it as a small, dedicated service pays off.
ReadSelected work
Things I've built
01
Agent Orchestrator
BuildingA policy-driven router that sits between a chat frontend and an LLM stack — classifies intent, dispatches to tools, workflows, or direct answers, and streams the result back over SSE.
02
Intent Classification Service
BuildingThe front-of-funnel decision for an LLM assistant — direct-answer vs clarify vs call-tool — using a live agent registry and structured LLM outputs.
03
Notification Fan-out Service
BuildingA durable event bus for agent platforms — authenticate, persist, and fan out to Slack, ticketing, email, and audit consumers via EventBridge + SQS.
04
MCP Servers
BuildingModel Context Protocol servers that expose internal capabilities — HR systems, knowledge search — as LLM-callable tools with OAuth 2.1 + PKCE.
05
n8n as an Agent Runtime
BuildingUsing a node-based workflow engine as the "tool provider" for an LLM agent — sub-workflows become tools the model can call.
06
Streaming AI Assistant
LiveA secure browser-based chat assistant — React + Server-Sent Events streaming from an LLM middleware over AWS Lambda, with per-user history in DynamoDB.
07
Internal Design System
BuildingA reusable React + JSS component library documented in Storybook — used across a fleet of internal tools to enforce consistency and accessibility.
08
Employee Support Portal
LiveA React 18 SPA for raising, tracking, and approving support tickets across multiple business groups — with contextual guided tours and knowledge-base search.
Say helloHave a question, an idea, or a small
Have a question, an idea, or a small
contract that needs a careful hand?
I'm always happy to talk about backend architecture, LLM stacks, and the tricky bits between them. Best reached over email.