I will build a multi agent ai assistant with langgraph that routes tasks
About this gig
One assistant, many specialists, and a router that sends each question to the right one.
I build multi-agent AI assistants (LangGraph) where a router decides which specialist agent handles a request, each agent uses its own tools and data, and anything out of scope is politely refused instead of guessed.
What you get:
- A router agent plus specialist agents with their own tools
- Shared memory and clean hand-offs between agents
- Refusal rules for off-topic or unsafe requests
- A test set so you can measure routing and answer quality
- A web chat interface plus an API for your app
Proof it works: on my 59 question test set the router picked the right agent 96% of the time (up from 79%), answers were correct 92% of the time (up from 75%), and 12 of 12 out-of-scope questions were refused.
Stack: Python, LangGraph, FastAPI, Next.js, TypeScript.
Message me first with the workflow you want to automate, and I will recommend the right package.
Get to know Moshiur Rahman
AI Full Stack Developer, NextJS, Python, Multi agent, Enterprise RAG
- FromBangladesh
- Member sinceOct 2021
- Avg. response time1 hour
Languages
Bengali, English, Hindi, French
My Portfolio
FAQ
What is a multi agent AI assistant?
One assistant made of several specialist agents. A router reads each request and sends it to the agent that owns the right tools or data, so answers are more accurate than a single general bot.
Can the agents use my own data and tools?
Yes. Each agent can be connected to your databases, documents or APIs. Tell me the workflow you want to automate and the systems involved, and I will confirm what is possible before you order.
What happens with questions the assistant should not answer?
They are refused politely instead of guessed. Refusal rules for off topic or unsafe requests are included in every package, and I test them on a question set.

