Building Block 4 - Connectors
A loop that can only see your files is a tiny loop. Connectors let it read your issue tracker, open pull requests, and ping your team
The Problem: A Loop in a Box
So far, your loop can read and write files on your computer. Useful - but limited. The real work of software lives in other places too: the issue tracker where bugs are logged, the database with live data, the chat channel where your team talks, the error monitor that screams when production breaks.
A loop trapped with only your files is like an employee who can read documents but can't use email, can't open a ticket, and can't tell anyone what they did. Connectors are how the loop reaches out and touches those real tools.
Connectors Run on MCP
Connectors are built on something you met back in Level 1: the Model Context Protocol (MCP). You don't need the deep technical details here - just the idea. MCP is a standard plug shape. Because every tool speaks the same "plug shape," a connector you set up in one AI tool usually works in another with little change.
What Connectors Unlock
This is the difference between an agent that describes a fix and a loop that actually does the whole job end to end:
WITHOUT connectors:
Agent: "Here is the fix. You should open a PR, update the
ticket, and let the team know." (you do the rest)
WITH connectors:
Loop: opens the pull request itself
โ links the Linear/Jira ticket and moves it to "In Review"
โ waits for CI to go green
โ posts "fixed the auth bug, PR #482" in Slack
โ leaves the hard cases in your triage inboxThe connectors are the reason the loop can act inside your actual workflow instead of just telling you what it would do if it could.
High-Value Connectors to Set Up First
Read repositories, create branches, open pull requests, and react to events like "a PR was opened." The backbone of most coding loops.
Update ticket status automatically and link pull requests back to the work item, so the tracker stays honest without anyone typing.
Post daily summaries and - importantly - alert a human when the loop hits something it shouldn't handle alone.
Watch for spikes in live errors and draft hotfixes for the most frequent ones, before a human even notices.
A Quiet Warning (We'll Return to It)
Connectors are also where a loop gains power in the real world - which means real consequences. A connector with write access can change tickets, push code, and message people. That's wonderful when it works and a problem when it doesn't.
Hands-On: Map Your Loop's Reach
Lets the loop reach beyond your files into real tools: trackers, databases, chat, error monitors
The "USB port" standard connectors run on - set up once, works across AI tools
Connectors turn "here is the fix" into a loop that actually opens the PR and updates the ticket
GitHub/GitLab, Linear/Jira, Slack/Discord, Sentry - the common coding-loop stack
Give a connector the minimum permission it needs; write access is power and risk
How much real-world damage a loop could do - track which connectors can write/change things