GitHub Now Catches Duplicate Issues Before Developers Even Submit Them
GitHub Issues now warns you about duplicate issues before you submit, and AI agents can fully triage issues via the MCP server
- Duplicate detection in preview: GitHub Issues now surfaces up to 3 matching issues inline as you fill out the creation form, before you submit.
- Real-time, not post-hoc: Suggestions appear while typing, letting reporters self-serve instead of waiting for a maintainer to close duplicates.
- MCP issue field support: AI agents connected to the GitHub MCP server can now read and write structured issue fields like Priority, Area, and Effort.
- Agentic triage unlocked: Agents can create fully triaged issues with all metadata set automatically, and filter the backlog by field values.
- Strong early adoption: Over 1,000 organizations have adopted issue fields since the initial preview, including large enterprises and OSS projects.
- Both features are free and available now in public preview; MCP field support requires the
remote_mcp_issue_fieldsflag (on by default in Insiders Mode).
If you maintain a popular open-source repo, you know the drill: someone files a bug, you search, find three existing issues saying the same thing, close it as a duplicate, and move on. Repeat, forever. GitHub is now taking a direct swing at this with two updates to GitHub Issues: real-time duplicate detection in the issue creation form, and full read/write access to issue fields via the GitHub MCP server.
Catch duplicates before they happen
Duplicate issues are one of the biggest time sinks for maintainers: triaging the same bug filed multiple ways, closing duplicates, and linking back to the original. For large repositories, this can take up hours every week.
As a first step to reduce maintainer triage time, issue creation now flags potential matches against existing issues in the repository as issue details are being populated. If potential matches are found, they appear inline in the issue creation form with up to three suggestions. You can review the suggested issues or continue creating your issue.
The key design choice here is timing: the check happens while you are filling out the form, not after you submit. That means the person filing the issue sees the suggestions and can self-serve, rather than a maintainer having to close it after the fact. The reporter can either link to the existing issue or decide their case is genuinely different and proceed.
What it is good at, and where it falls short
The feature shines in large, active repositories where the same bug gets filed repeatedly by different users with different wording. Because the matching runs on issue content as it is typed, it can catch semantic near-duplicates rather than just exact title matches. That said, GitHub has not published details on the underlying similarity model, so the quality of suggestions on niche or domain-specific repositories remains to be seen in practice.
- Good for: Popular OSS projects, large enterprise monorepos, any repo with a high volume of external contributors
- Less useful for: Small private repos with few issues, or repos where issues are highly technical and unlikely to be duplicated verbatim
- Limitation: Suggestions cap at three matches, and there is no automatic closing or labeling; a human still decides
This feature is available as a public preview. Feedback in the community discussion directly shapes what GitHub builds next.
AI agents can now fully triage issues
The second part of this release is aimed squarely at teams building agentic workflows. AI tools connected to the GitHub MCP server can now read and write issue fields. Agents can create fully triaged issues with priority, area, dates, and other fields automatically set, plus filter existing issues by field values.
MCP (Model Context Protocol) is an open standard that lets AI agents talk to external tools and services through a unified interface. Think of it as a plugin system for AI agents: instead of writing custom API integrations, you connect your agent to an MCP server and it gains structured access to that platform's capabilities.
Issue field support is now available behind the remote_mcp_issue_fields feature flag, and is enabled automatically in Insiders Mode. This includes listing issue fields, filtering issues by custom field values, and reading or writing issue field values.
This matters because issue fields, which were themselves only recently opened to all organizations, are the structured metadata layer sitting on top of every issue. When you define typed metadata like Priority, Effort, or any custom field at the org level, it automatically shows up on every issue, in every repository. Now an agent can set those fields programmatically, not just a human clicking through a form.
What this unlocks in practice
Combine both features and the picture becomes clearer. A Copilot agent or any MCP-connected tool can now:
- Receive a bug report from Slack, a webhook, or a CI failure
- Check for existing duplicates before creating a new issue
- Create a fully triaged issue with
Priority,Area, andEffortfields already populated - Filter and query the issue backlog by those field values for reporting or planning
The GitHub MCP Server connects AI tools directly to GitHub's platform, giving AI agents, assistants, and chatbots the ability to read repositories and code files, manage issues and PRs, analyze code, and automate workflows, all through natural language interactions.
Early adoption of issue fields has been strong. Over 1,000 organizations have adopted issue fields since the initial preview, including some of the largest enterprises and open source projects on GitHub. Teams describe the combination of structured fields and automation as the missing link between issues and projects.
How to get started
Duplicate detection requires no configuration. Open any repository, start creating an issue, and the suggestions will appear inline as you type. For the MCP side, you need the GitHub MCP server connected to your agent host. Compatible hosts include VS Code 1.101+, Claude Desktop, Cursor, and Windsurf. Issue field support is gated behind the remote_mcp_issue_fields flag but is on by default in Insiders Mode.
Both features are free and available now in public preview on github.com. The duplicate detection is the more immediately useful of the two for most teams, but the MCP issue fields update is the one that will matter most as agentic workflows become a standard part of how issues get created and managed.