How RepoHelpdesk works
A team connects a repository, and everyone in the project can ask questions about it in plain English — answered by AI actually reading the code.
Connect your repo
An admin links a repository and picks a branch. The Agent keeps it in sync wherever you run it — your teammates never need a copy of the code.
Invite your team
Add teammates by email. They get to ask questions about the codebase — but they only ever receive answers, never the source files, secrets, or keys.
Chat with the code
Anyone in the project asks in plain English — "Where is authentication handled?" — and the answer streams back live, read straight from the actual code.
The problem it solves
Most of what a team knows about its code lives in just a few people's heads. When someone needs to know how part of the app works, they either dig through unfamiliar files for hours or interrupt the person who wrote it. The answer is in the code — it's just hard to get to.
RepoHelpdesk fixes that. You connect your code once, and anyone on the team can simply ask a question in plain English — like "Where is login handled?" — and get a clear answer, drawn straight from your actual code. No digging, no waiting on the one person who knows.
Why not just hand out the code?
You could use a regular tool like Claude Code or ChatGPT instead. But to answer questions about your code, those tools need a copy of the code first — on someone's laptop, pasted into a chat, or connected to a tool they control. Once that happens, your code and its passwords and keys are in their hands. With RepoHelpdesk, they never have to leave your control.
- Each person needs their own copy of the code to use it.
- Anyone who can ask a question can also read it, copy it, and walk away with it.
- Any passwords and keys saved in the code come along too.
- Your code stays where you put it — only RepoHelpdesk reads it, on your team's behalf.
- Teammates get answers, not files — they can ask anything without ever seeing the code.
- Passwords and keys stay hidden — they're never shown in an answer.
The three roles
The architecture cleanly separates three roles — and that separation is exactly what makes its deployment flexibility possible. The Repo Agent's logic is identical everywhere; only its host changes.
The browser. Types questions and renders the streamed answers.
Auth, projects, members, invites, chat history, usage reporting, and message routing.
Holds the repo files, calls the AI model, and runs the tool loop. The "brain."
How a question travels
A question can start wherever your team already works — the RepoHelpdesk app, the RepoHelpdesk chat bot in your own app, Jira, Slack, or Microsoft Teams. Wherever it begins, it follows the same simple path, and the answer comes back the same way.
The Relay passes the question to the Agent inside your environment, which reads your continuously synced code and sends the answer back — returned right where it was asked.
Deployment modes
The same product ships three ways, differing only in where the Repo Agent runs.
| Mode | Files live on | Install | Best for |
|---|---|---|---|
| Self-hosted Agent | Your server | repohelpdesk-agent |
Regulated orgs — code can't leave the network |
| Managed Agent | A dedicated server we run for you | None | Teams that would rather we handle the hosting — nothing to run yourself (any plan) |
Whichever you choose, the Relay never sees your repo files — it only passes messages along. The Agent does the reading, wherever it runs: on your server, or on a server we run for you.