Skip to content

Strategy for integrating (or not) third-party MCP servers, e.g. Sentry #6706

Description

@jpetto

The Homepage New Tab team is looking to have an agent able to triage Sentry alerts in order to reduce time to root cause/remediation.

The strategy in question is - do we want to write our own MCP server (giving agents a limited set of tools backed by API integrations with the third-party), or do we want to proxy official third-party MCP servers?

Assumptions here are:

  • We want to continue preventing the agent from accessing auth backed services directly (sidecar/proxy)
  • We want explicit control over what actions are made available to the agent to perform its task

WIP PR towards a custom MCP for Sentry is here.

Metadata

Metadata

Assignees

No one assigned

    Labels

    Fields

    Priority

    None yet

    Start date

    None yet

    Target date

    None yet

    Effort

    None yet

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions