Skip to main content

GitHub

There are two ways to use GitHub with Revelry Ai, and they serve different purposes:

  • GitHub MCP tools let the AI read and work with repositories, issues, and pull requests in chat.
  • The GitHub data source imports repository content (code, issues, and docs) into Knowledge as searchable documents for chats and artifacts.

The two connections are independent. You can use one or both.

What it can do​

With GitHub MCP tools enabled in chat, you can ask the AI to:

  • Search and browse repository code
  • Look up issues and pull requests
  • Review PR diffs and provide feedback
  • Create or comment on issues
  • Explore repository structure and documentation

Connecting a repo as a data source imports its content into Knowledge as searchable documents. The AI can then use your code, issues, and docs as context in any chat.

Before you start​

  • Both connections need a GitHub personal access token. See Access tokens.
  • To connect a data source, you need permission to manage documents, and your plan must include data connectors.
note

GitHub MCP tools are off by default. Contact your Revelry Ai team to enable.

GitHub MCP tools​

GitHub MCP tools let the AI work with your repositories during a chat. It can look up issues, review pull requests, and search code.

Personal vs. shared configuration​

You can configure GitHub MCP at two levels. The organization's shared configuration always takes precedence over a personal one.

Shared (organization-level)​

An admin configures it under Integrations > Shared. Once it's set, every user in the organization who uses GitHub MCP tools uses this token.

For admins

Configure the organization-level GitHub MCP only if every user should share one GitHub connection. If team members need different repositories, or need to act as themselves on GitHub, leave it empty so each user can add a personal token.

Personal​

Each user configures it under Integrations > Personal with their own personal access token, so their GitHub actions happen under their own identity.

A personal configuration is active only when the organization has no shared GitHub configuration. A shared token overrides every personal one.

Setting up GitHub MCP​

  1. Open Integrations, under My Organization in your account menu at the bottom of the left navigation
  2. Go to the Personal tab, or the Shared tab if you manage integrations for your organization
  3. Find GitHub and click Connect, then click Connect MCP Tool
  4. Enter a GitHub personal access token (see token setup below), choose the toolsets the AI can use, and click Save Configuration
  5. Turn on GitHub tools in chat from the Chat Tools menu

GitHub data source​

Creating the connection​

To connect a repo, go to Knowledge > Sources, click New Data Source, and choose GitHub Repository.

GitHub URL​

The URL of the repo you want to connect. Revelry Ai checks the URL and token with GitHub when you click Connect.

Project​

Choose Organization-wide (the default) to make the repo available to every project, or pick one project.

Access token​

This is the access token you generated. See token setup below. When you edit the repo later, the field is New Token (optional).

Sync status​

Revelry Ai syncs the repo, including its issues and files, when you save the connection and again every day at 5:00 UTC. The data source's sync status shows whether that import worked. It doesn't depend on GitHub MCP, so the data source can sync with or without MCP tools connected.

Troubleshooting​

If you don't own the repository, GitHub may ask an organization owner to approve a fine-grained token. If it does, you can't connect the repo to Revelry Ai until they approve it.

Limitations​

  • The data source doesn't sync pull requests. GitHub MCP tools can read them.
  • The data source syncs when you save the connection and once a day, so recent changes can take up to a day to appear.

Access tokens​

GitHub MCP and the GitHub data source both need a personal access token. Use a fine-grained personal access token. Classic tokens also work. See the GitHub documentation for how to create one.

Fine-grained personal access tokens​

A fine-grained token lets you set an expiration date and choose which repos it can reach.

You also choose the token's repository permissions. Start with the ones below, and add more, such as Projects, as you need them.

PermissionAccess levelDescription
Contentsread-onlyLets Revelry Ai read and index your repository's files.
Issuesread and writeLets Revelry Ai index existing issues and export generated issues to GitHub.
Metadataread-onlyGitHub requires it for the other permissions.
Pull Requestsread-onlyLets GitHub MCP tools read pull requests. The data source doesn't sync pull requests.

Classic access tokens​

Classic tokens can't be limited to specific repositories or permissions. Grant the full repo scope, and Revelry Ai uses the same operations listed for fine-grained tokens.