AI & Automation
Agent Memory: When It Helps and When It Creates Risk
Agent memory speeds up repeated tasks and quietly turns stale facts into confident answers. A practical look at when to use it and how to bound it.
Tag
ai agents
9 posts
AI & Automation
Connecting an AI Agent to Slack, GitHub, Wikis, and Tickets
A practical pattern for connecting an agent to four different systems: one shared connector interface, so search and citations don't special-case each source.
AI & Automation
MCP Security: Permissions, Secrets, and Tool Boundaries
A practical MCP security model: scope permissions per caller, keep secrets out of the model's context entirely, and never let a tool exceed its agent's access.
AI & Automation
How to Prevent an AI Agent From Taking Unsafe Actions
Preventing unsafe AI agent actions is an architecture problem, not a prompting one: four concrete layers that actually stop damage before it happens.
AI & Automation
Designing MCP Tools for Business Systems
Practical rules for designing MCP tools that an agent can actually use well: narrow scope, task-shaped names, bounded results, and no mode flags.
AI & Automation
RAG Versus Fine-Tuning for Operational Knowledge
Why fine-tuning is the wrong tool for knowledge that changes weekly, and when it's actually the right one: a practical comparison for business teams.
AI & Automation
Domain-Specific AI Agents: Why Composition Beats One Big General Agent
Why the next wave of AI automation is many narrow, domain-specific agents instead of one general-purpose agent, and how to start building that way.
AI & Automation
What I Learned Building a Citation-Based RAG System
Six lessons from building Nexus's citation system: why similar text isn't trustworthy evidence, and what actually makes a RAG answer verifiable.
AI & Automation
How to Build Your Own AI Command Center: A Self-Hosted MCP Server
"AI is amazing, but existing solutions limit your control... here's how to build your own."