Skip to main content
Unblocked MCP tools are actively evolving. Tools and their descriptions may change over time.
unblocked_context_engine is a focused lookup that quickly returns content related to a specific question — especially when the answer to “why does this exist?” or “why is it done this way?” isn’t visible in the code itself. It pulls relevant PR discussions, past decisions, team conventions, and documentation for one topic at a time. Code search shows you what the code does now. The context engine adds:
  • Why it was built this way — PR discussions, design decisions, and review comments that explain the reasoning behind implementation choices.
  • What was tried before — rejected approaches, prior incidents, and previous implementations that didn’t work out.
  • What the team expects — conventions from documentation, messaging threads, and review feedback.
  • What’s documented elsewhere — related issues, documentation pages, and team discussions that provide context the code alone doesn’t show.
If your question is purely about current implementation, your agent uses code search directly. When the question involves intent, history, or conventions, the context engine surfaces organizational knowledge that isn’t in the code.

​
When your agent uses it

  • When it needs to understand what a piece of code does and why it exists that way, such as why a payment processor retries three times instead of using exponential backoff.
  • Before finalizing a non-trivial plan, to check for prior decisions or conventions, such as whether the team has an existing retry utility.
  • When it has a focused question about a specific class, service, endpoint, or feature.

​
When your agent switches to another tool

  • If the question turns out to span multiple systems or teams, it may escalate to research_task.
  • If it needs to query specific records by filter (author, status, date), it will use data_retrieval.