Web retrieval
Fetch
Fetch and convert web content for model-readable context.
Score breakdown
Evidence history
Review snapshots show why the current score, confidence, and risk labels changed over time.
| Date | Change | Score | Confidence | Risk |
|---|---|---|---|---|
| May 12, 2026 | Initial MCP Rank review captured from the curated seed dataset. Narrow surface area and clear usefulness. seed snapshot · reviewed | 79 | high | low |
Evidence notes
Best-fit use cases
Best for source-grounded answers that need public documentation, changelogs, or release notes.
Useful for review workflows where the assistant must cite exact public pages before making recommendations.
A good companion to search tools: search finds candidates, Fetch retrieves the specific source text.
Risk analysis
Fetched content is untrusted input and can contain prompt-injection instructions aimed at the agent.
Internal URL fetching can expose private services if network policy is loose.
Default public-fetch usage is lower risk than local file or browser automation access.