Skip to main content
Unblocked MCP tools are actively evolving. Tools and their descriptions may change over time.
context_research runs a multi-source research query across code, pull requests, documentation, issues, and messaging. It returns synthesized context with citations, so your agent can understand not just what the code does today but why it was built that way, what patterns exist across the codebase, and what the team has already decided. This tool has the same functionality as the unblocked context-research CLI command — the MCP surface and the CLI surface are backed by the same engine.

​
What it returns

  • Summary — a synthesized answer that pulls together what was found across sources.
  • Sources — the specific PRs, documents, issues, messages, and code snippets that back the summary, each with a title, URL, and relevant content.
Use the effort input to trade latency for thoroughness; higher effort runs a more exhaustive search across your connected sources.

​
When your agent uses it

  • Investigating unfamiliar subsystems or feature areas, such as how the notification pipeline works end-to-end.
  • Understanding why a piece of code exists the way it does — PR discussions, past decisions, and review comments that explain implementation choices.
  • Finding specific records by filter: issues, pull requests, messages, or threads scoped by author, status, repo, channel, or date range.
  • Scoping an integration, migration, or refactor, such as evaluating the impact of upgrading a major dependency.
  • Questions that span multiple systems, teams, or sources where no single lookup will give the full picture.
  • At the start of a planning task, in parallel with code exploration.

​
When your agent uses something else

  • For a known URL that just needs to be fetched, context_get_urls resolves the content directly.