AuricIDE / Blog
Blog
Notes from building AuricIDE: agent orchestration, machine-checkable goals, and systems for supervising coding agents.
-
AI Test Case Generation: Write the Test Before the Code
Learn how AI test case generation works on a ticket, and why the test must reach the agent before it writes code.
-
MCP Use Cases: When an MCP Tool Earns Its Place
MCP use cases that justify a tool: shared project state and reaching a person. Everything else can stay a file, and here is how to tell.
-
AI Agent Management Software: Where Handoffs Break
Compare AI agent management software on orchestration, integrations, local execution and team workflows, with a worked example from AuricIDE.
-
Git Worktree Explained: Parallel Branches Without Clones
Git worktree lets you check out several branches at once, each in its own folder. Learn how it works, how it differs from a clone, and how to set one up.
-
MCP Project Management: How Agents Take One Ticket
Learn how MCP project management lets AI agents take one ticket from a shared backlog without double claims, and what that design costs.
-
MCP vs API: Why a Model Needs a Different Interface
See how one MCP vs API call plays out in auric-pm, why tools are written for a model caller, and when a plain API fits better.
-
Coding Agents: Where the Record Should Live
Where should an agent handover be recorded: git history, a pull request, the editor session or a shared project store? The trade-offs, and a test you can run.
-
Agent Swarm Explained: Coordination, Tasks and Pitfalls
Agent swarm basics for engineers: coordination patterns, suitable tasks, handoff mechanics, and the trade-offs to weigh before you adopt one.
-
Single Agent vs Multi Agent: Which Coding Setup Fits
Compare single agent and multi agent coding setups by task fit and trial results, with a rule for choosing one before you launch a fleet.
-
Prompt Chaining Explained: Why Handoffs Need Checks
Prompt chaining passes a cleaned tail of each step's output forward. Learn why the next agent should verify that handover before building on it.
-
AI Requirements Management: One ID Per Requirement
See how AI requirements management works in AuricIDE: one record per requirement, a unique short ID, and a batch import that fails whole.
-
Agent Supervisor: Which Limits Belong in the App
An agent supervisor in AuricIDE follows two kinds of rules: fences the app checks on every spawn, and notes the model reads and has to judge.