---
title: "Replacing ClickUp with GitHub Issues in One Month"
url: https://www.insulin.dev/blog/replace-clickup-with-github-issues/
canonical: https://www.insulin.dev/blog/replace-clickup-with-github-issues/
type: Blog
description: "How we replaced ClickUp with GitHub Issues in a month: the architecture behind our customer and staff ticket consoles, two demo videos, and the lessons."
---

# Replacing ClickUp with GitHub Issues in One Month

> Canonical HTML version: https://www.insulin.dev/blog/replace-clickup-with-github-issues/

1.  [Home](/)
2.  [Blog](/blog/)
3.  How We Replaced ClickUp with GitHub Issues in One Month

# How We Replaced ClickUp with GitHub Issues in One Month

We moved every engineering task and customer-support ticket from ClickUp to GitHub Issues, behind two consoles we built ourselves. One week of engineering, two to three weeks of testing and migration, and about $15,000 a year saved.

![Shirley Guo](/authors/shirley-guo.jpg)

Shirley Guo

Nov 11, 2026

 ![How We Replaced ClickUp with GitHub Issues in One Month](/images/blog/replace-clickup-with-github-issues/hero.png)

In September 2026 we moved every ticket we work, from engineering tasks and product work to customer-support requests, out of ClickUp and onto GitHub Issues. Customers still file and follow their tickets in the Fours console. Our team works them in an internal console. AI agents work the same issues with GitHub’s `gh` command line. The first version merged on September 10 after one week of engineering, and testing, migrating the old tickets and polishing took another two to three weeks, run as a side project next to everything else. It saves us about $15,000 a year.

This post covers why we left, how the pieces fit together, what it looks like in two short demo videos (sample data), and what we would tell a team weighing the same move.

## Why did we leave ClickUp?

**We were paying for a full work-management suite and using its ticket list.** ClickUp does a great deal. We used two things: tasks for product and engineering work, and tickets for customer support. Three reasons made that slice worth rebuilding.

-   **Every tool is one more endpoint for an AI agent.** Agents now help us design, build and support, and every system they touch adds credentials, an API, a permission model and rate limits to keep working. Moving tickets onto GitHub, where the code already is, took one system off that list.
-   **Tickets belong next to the code.** Our repositories and pull requests are on GitHub. A bug report in another tool is a link someone copies across by hand; an issue in the same place is something a pull request can point at.
-   **It cost about $15,000 a year.** GitHub Issues came with the GitHub plan we already pay for.

## Why GitHub Issues, and not another tracker?

**It added no new bill, and our AI agents already knew how to use it.** We judged the options on four criteria.

| Criterion | What we needed | What GitHub Issues gave us |
| --- | --- | --- |
| Cost | Nothing new to buy | Included in the GitHub plan we already had |
| AI agents | An interface agents handle without special setup | The `gh` command line, plain Markdown, and issues next to the code |
| Link to the work | A ticket tied to the pull requests that fix it | Pull requests and issues in the same place |
| Room to build | A documented API to put our own screens on top | A public API for issues and Project boards |

What GitHub Issues does not do is serve customers. A customer should not need a GitHub account, and must never see another company’s tickets. That gap is the part we built.

## How is the ticket system put together?

**Two consoles share one ticket service in front of GitHub, and AI agents go to GitHub directly.** GitHub is the source of truth; everything else is a way in.

![Architecture of the ticket system. Customers use the Fours console and our CS, engineers, PMs and QA use the internal console; both call one ticket service, which keeps a ticket index, access rules and email and Slack notifications, and syncs both ways with GitHub. AI agents use the gh CLI on the same GitHub issues directly. On GitHub, each ticket is one issue, its status is a column on a Project board, and pull requests link to the issue they fix.](/images/blog/replace-clickup-with-github-issues/architecture.png)

-   **The Fours console** is where customers file a ticket, follow its status and reply. Each company sees only its own tickets.
-   **The internal console** is where our customer success team, engineers, product managers and QA work every customer’s tickets in one queue: as a table or a Kanban board, by sprint, grouped by owner, with saved views.
-   **AI agents use `gh`** on the same issues: they read and triage tickets, comment, and open the pull requests that fix them.
-   **The ticket service** behind both consoles keeps a small index of what we filter on (company, priority, assignee, sprint, product manager), decides who may see which ticket, and sends email and Slack notifications.

Each ticket is one GitHub issue. Its description and comments live only in the issue, and both consoles read them live. Its status is a column on a GitHub Project board, and the columns are the same statuses the consoles show: Open, In Progress, In Review, Pending Deployment, Waiting on Customer, Done and a few more. A change made in either console is pushed to GitHub, and a change made on GitHub is swept back into the index.

## What does it look like in practice?

### A customer files and follows a ticket in the Fours console

Sample data. A customer at the fictional Acme Analytics files a ticket, then replies to our team on an older one. The customer never sees GitHub.

### Our team triages it and runs the sprint in the internal console

Sample data. The new Acme ticket arrives in the staff queue, where it is assigned, labelled and answered; then a card moves to In Review on the sprint board.

## How did it fit in one month?

**We built only what we used, and kept the API the customer console already called.**

-   **Week one: engineering.** We put GitHub Issues behind the ticket API the Fours console was already calling, in place of the ClickUp client, and built the staff queue in the internal console.
-   **Weeks two to four: testing, migration and polish.** Existing ClickUp tickets were migrated with their original authors and dates, so an old ticket’s history still reads correctly. Then we fixed what daily use turned up.
-   **As a side project.** Nobody stopped other work for it, and the consoles have kept improving with daily use since.

We later did the same for our equity records: [how we replaced Pulley with an open-source cap table](/blog/replace-pulley-with-open-source-cap-table/).

## What would we tell a team making the same move?

1.  **Decide once where each fact lives.** Ticket text lives in the issue and is read live; our index holds only what we filter and sort on. Two copies of a description drift apart.
2.  **Assume people will edit on GitHub.** Our first design assumed nobody would change the issues on github.com. They did: people and automation close and reassign hundreds of support issues a week there. Sync in both directions from the start.
3.  **Keep customers out of GitHub.** Your service, not GitHub’s permissions, decides which company sees which ticket. That is most of the reason for a customer console.
4.  **Use one set of statuses everywhere.** The same names in the consoles and on the Project board mean a card moved anywhere means the same thing everywhere.
5.  **Give agents the interface they already use.** Our agents already worked with `gh` for code and pull requests, so tickets in the same place gave them nothing new to learn and no new credentials to hold.

## What did we give up?

**The features we never used, and the convenience of features we may want later.** ClickUp offers far more than tickets, and dropping what we did not use cost us nothing we could name. Anything new, such as service-level reports for support, is now ours to build rather than a setting to switch on. And the ticket service now depends on GitHub’s availability and API limits, which a hosted tool would have absorbed for us.

## Where Insulin fits

If your team keeps its tickets in GitHub too, an Insulin agent can answer from them: [connect a repository to a knowledge base](/blog/connect-external-sources-to-a-knowledge-base/) and its issues and pull requests sync in as Markdown, without the source code.
