LEPA Support Agent
A LangGraph support workflow for LEPA: typed state, conditional routing, tools, memory, and a multi-agent split behind FastAPI.
- Python
- LangGraph
- FastAPI
- SQLite
- OpenRouter
Agent Harness
A small TypeScript runtime that connects an LLM to remote DevAssist MCP tools and prints every step of the agent loop.
Overview
The agent harness is a TypeScript app that orchestrates an LLM plus remote DevAssist tools so the agent loop is visible. It loads config, connects to MCP over HTTPS, passes tools into generateText, prints each step, enforces a max-step limit, and closes the session. It intentionally avoids larger agent frameworks.
Problem
After shipping local tool calling and a remote MCP server, I still needed to see how an application — not just Cursor — runs an agent against those tools.
What I built
A thin harness connects to the deployed DevAssist endpoint, exposes tools to the model, executes tool calls over HTTPS, feeds results back, and stops safely. No LangChain, LangGraph, or similar orchestration libraries.
My contribution
- Built a readable agent loop over remote MCP with step logging, limits, and session cleanup.
Technologies
- TypeScript
- AI SDK
- MCP
- Node.js
More ai / llm
AI Document Processing Pipeline
An async document pipeline: upload or webhook, Temporal extract/analyze/summarize, then Trigger.dev email — without blocking the request.
- Temporal
- Trigger.dev
- TypeScript
- Webhooks
- LLM
AI Job Application Assistant
Parse a job and a CV, explain the match, draft materials, then stop for a human. The AI never applies.
- Python
- LangGraph
- FastAPI
- Human-in-the-loop
- Evaluation
Related writing
I Built an AI Agent Harness That Uses Remote MCP Tools
How I connected an AI SDK agent to my deployed MCP server and finally understood what the harness around an agent actually does.