Skip to main content
Unblocked MCP tools are actively evolving. Tools and their descriptions may change over time.
data_retrieval retrieves records — not guidance or explanations, but the actual items themselves: issues, pull requests, messages, and threads. It returns results with lower latency than research_task because it queries your integrated services directly rather than synthesizing across them. Your agent uses it when it has a specific identifier or filter (a repo, author, status, date range, channel, or issue key) and needs to retrieve matching results. For “how to” questions, best practices, or exploratory questions, your agent uses unblocked_context_engine or research_task instead.

​
What it retrieves

  • Single issues by key
  • Filtered lists by project, status, assignee, or date range
  • Epics, sprints, and board contents
  • Issues by label or saved filter

​
When your agent uses it

  • Listing or summarizing pull requests by author or status, such as your open PRs across all repos.
  • Finding recent commits or checking CI/build status for a PR.
  • Summarizing a messaging channel or thread.
  • Looking up issues, epics, or sprint work in your issue tracker, such as what’s in the current sprint for the Platform board.
  • Checking what someone worked on during a given time period.
  • Finding who is the expert on a specific file or topic.

​
When your agent uses something else

  • For understanding why code exists the way it does, unblocked_context_engine provides the surrounding context.
  • For broad investigations that span multiple systems, research_task synthesizes a full briefing.