GitHub in your team chat, read by AI

How Waffle works with GitHub

Connect GitHub once and Waffle AI works with your code from chat. It browses repositories, reads files, searches code, and reads pull requests, and with an admin opt in it opens a draft pull request that a human approves before merge. All of it happens in the channel, where knowledge workers already toggle between applications about 1,200 times a day Harvard Business Review, 2022.

The idea is simple: instead of leaving chat to go dig through the repository, you ask in the channel and the answer comes back with the file and the lines that back it up.

What Waffle AI does with GitHub

Waffle AI does not just point at GitHub. It reads and reasons about what it finds, then answers with specifics.

• Reads your code and answers with file level specifics, not vague summaries. Ask why a function behaves a certain way and it points to the file and the lines.
• Explains a failing deploy from the commit that caused it, so the person who broke the build and the person debugging it are looking at the same evidence.
• Searches the codebase across a repository and returns the files that match, so you can find where something lives without leaving the channel.
• Reads a pull request, including its description and the per file diff, and summarizes what changed and why when anyone asks.
• Opens a draft pull request that an admin approves before it can merge, so the AI can propose a change without ever shipping one on its own.

That matters because the alternative is expensive. McKinsey estimated that employees spend about 1.8 hours a day searching and gathering information McKinsey Global Institute, 2012. Answering a code question where the conversation already lives cuts that hunt short.

Not just notifications

A typical Slack GitHub app posts commit and pull request notifications into a channel. That is useful, but it is a feed, not an assistant. Waffle AI reads and reasons about your code: it explains a failing deploy from the commit that caused it, searches the codebase, and reads a pull request diff, then answers in plain language. As of 2026 the difference is the reasoning, not the feed.

Bringing that reasoning into chat fits how people already work. In the 2024 Work Trend Index, Microsoft and LinkedIn reported that 75 percent of knowledge workers use AI at work Microsoft and LinkedIn Work Trend Index, 2024. Waffle AI puts that help next to your GitHub activity instead of in a separate tab.

Admin control and security

Access to GitHub is something an admin grants and controls. Connecting a repository grants read access only. Writes are opt in per connector, and any pull request the AI opens is a draft that GitHub cannot merge until a person marks it ready. The controls are enforced by the platform, not just by policy.

• Per channel connector scoping. An admin decides which channels can reach GitHub, so a public channel does not expose your repositories.
• Read only by default, with read and write enabled per connector. Waffle AI can read your code the moment you connect, and only make changes after an admin opts in.
• Connector tokens are AES-256-GCM encrypted at rest and never shown again after you save them.
• Draft only pull requests. A pull request the AI opens is a draft, which GitHub cannot merge until a human marks it ready, so the approval step is enforced by GitHub itself.

Frequently asked questions

Can Waffle AI read my private GitHub repositories?

Yes, once an admin connects GitHub for your team. Waffle AI gets read access to the repositories you connect: it can browse directories, read files, search code, and read pull requests. Read access is the default. Writes are opt in per connector, and any pull request the AI opens is a draft that an admin must approve before it can merge.

Can Waffle open pull requests?

Yes, but only as a draft, and only when an admin has turned on write access for the GitHub connector. Waffle AI opens the draft pull request inside the channel with a real branch and diff. GitHub will not let a draft merge until a person marks it ready for review, so an admin always approves before anything ships.

Is Waffle a GitHub bot?

No. A typical GitHub bot posts notifications about commits and pull requests into a channel. Waffle AI is a member of your chat that reads and reasons about your code: it explains a failing deploy from the commit that caused it, searches your codebase, and reads a pull request diff. You mention it like a teammate and its reply lands once, finished.

Does connecting GitHub let Waffle AI change my code without approval?

No. Connecting GitHub grants read access only. Making changes requires an admin to enable write access for that connector, and even then Waffle AI can only open a draft pull request. Because GitHub blocks a draft from merging, a human reviews and approves every change before it lands.

Can I limit which channels can use GitHub?

Yes. GitHub, like every connector, is scoped per channel by an admin. You can allow it in your engineering channels and leave it off everywhere else, so the AI only reaches your repositories where you intend it to.

Sources

The statistics cited on this page, with their original sources:

• Harvard Business Review, How Much Time and Energy Do We Waste Toggling Between Applications? (2022)
• McKinsey Global Institute, The Social Economy: Unlocking Value and Productivity Through Social Technologies (2012)
• Microsoft and LinkedIn, 2024 Work Trend Index Annual Report (2024)