Waffle for engineering teams

Your codebase, in the conversation

Connect GitHub and Waffle AI browses directories, reads files, searches code, and reads pull requests, and it can open a draft pull request that an admin approves before it can merge. Engineering questions get answers with file level specifics right where you are already talking, so nobody leaves the channel to go dig.

Waffle AI is a real member of your team chat, not a separate tab. You mention it like any teammate and its reply lands once, finished, with any files attached. When it reads your repositories it points at the exact paths, so a question about how something works comes back grounded in the actual source instead of a guess.

What Waffle AI does for engineers

• Reads your code and answers with file level specifics, not vague summaries.
• Explains a failing deploy by reading the commits and pull requests behind it.
• Opens a draft pull request for an admin to review and approve before it can merge.
• Files issues in Linear or Jira through MCP connectors, right from the thread.
• Writes runbooks, specs, and postmortems as real Word, PDF, or Markdown files.
• Catches you up on a channel with a private summary that never posts publicly.

Voice and screen share for pairing

Every channel can be a one click voice channel with screen share, so pairing on a bug or walking through a change is one tap away, in the same place your team already chats. No scheduling, no separate call app, and no link to send.

Voice channels carry speaking rings on avatars, push to talk, mute and deafen, and per person playback volume. Share your screen to review a diff together, then drop back into the thread when you are done. It is the low friction voice of a community app, tuned for how engineers actually work.

Admin control engineers trust

Admins decide exactly what Waffle AI can reach. Connectors are scoped per channel, each connector can be set read only or read and write, tokens are encrypted at rest, and private channels plus role based access keep sensitive work closed. The AI is a scoped colleague, never a loose cannon.

• Per channel connector scoping: all, none, or a hand picked set.
• Read only or read and write, chosen per connector.
• Connector tokens encrypted at rest with AES-256-GCM, never shown again after setup.
• Owner and admin roles with granular permissions, plus private channels.

Frequently asked questions

Can Waffle AI read my private GitHub repos?

Yes. Once an admin connects GitHub, Waffle AI can browse directories, read files, search code, and read pull requests in the connected repositories with read access. Writing is opt in and off by default. When an admin enables it, the AI can only open a draft pull request, which stays unmergeable until an admin approves it.

Can Waffle open pull requests?

It can open a draft pull request when an admin has turned on write access for GitHub. The pull request lands as a draft, so it cannot merge until an admin reviews and approves it from the in chat approval card. Rejecting closes the pull request. Any member can ask Waffle AI for a plain summary of the diff.

Does Waffle work with Linear and Jira?

Yes, through MCP connectors. An admin connects Linear or Atlassian (Jira and Confluence), or any custom MCP server, and can allow read only or read and write per connector. Admins also choose which connectors each channel may use. Waffle AI can then read and file issues in the tools your engineers already work in.

Is it safe to connect our tools?

Connector tokens are encrypted at rest with AES-256-GCM and are never shown again after setup. Anything a connected tool returns is treated as untrusted input and never as instructions. Access is scoped per channel and per connector, writes are opt in, and roles and private channels keep sensitive spaces closed.