A tool-using agent should not turn an unsupported field into a plausible answer. Return an evidence envelope instead of a bare value: show what each claim is based on, when the source was checked, and which requested details remain unknown. The envelope is an application-level response design—not a feature built into MCP.
Why a bare tool response invites guesses
If a tool returns only "active", the agent has no structured indication of which source supports that status, when it was retrieved, or whether the user’s other question—“what is its legal name?”—was answered at all. The model may fill in those gaps with a plausible completion. Plausibility is not evidence.
As an Amazon Associate I earn from qualifying purchases.
An evidence-linked response makes both the support and the gaps explicit. The agent can tell the user that the supplier is listed as active, identify the source and retrieval time, and say “I don’t know” about its legal name if no source supports that field. As Baron Sigma puts it in the FACTRAIL tutorial, “The design rule behind it: unknown is better than invented.”
What goes in an evidence envelope?
The FACTRAIL tutorial describes schema 1.2 as an envelope with several related parts. Together, they let an application return a result and the context needed to interpret it.
#1 Best Overall
- More for the money with this high quality Product
- Offers premium quality at outstanding saving
- Excellent product
- 100% satisfaction
status: the overall result, such as supported, contradicted, insufficient evidence, stale, or conflicting sources.facts[]: individual values, their support levels, and the IDs of evidence that support them.evidence[]: source or publisher, URL, retrieval time, and source status.coverage: which requested fields were resolved, which were not, and why.conflicts[]: competing claims retained for the agent to report rather than silently choosing between them.freshness: information about retrieval and staleness.receipt_id: an identifier for retrieving the same envelope again and checking whether its content has changed.
This is a design pattern, not a guarantee that a source is correct or that a tool has checked every relevant source. The application must define what its support levels mean and how it gathers, timestamps, and preserves evidence.
How to handle partial evidence without guessing
A top-level status does not have to mean every field has the same outcome. In the tutorial’s sample EDF company result, five fields are supported, while legal_form is unsupported. The overall status is insufficient_evidence. That status preserves an important distinction: some answers are supported, but the requested record is incomplete.
Have the agent report the fields separately. For example, it can state the supported facts with their sources and explain that the legal form was not established by the available evidence. It should not infer a legal form from the company name, activity, or another field. Preserve unresolved values as unresolved in the tool result as well as in the final answer; otherwise, later steps in an agent workflow may mistake a guess for a verified fact.
FACTRAIL’s example scope and limits
The FACTRAIL tutorial describes two capabilities: company_fr, a verification tool for French companies using INSEE Sirene and BODACC, and an EU import assessment described as beta. The tutorial says official EU TARIC data was not installed at the time it was written. These are service-specific details reported by that tutorial, not general properties of MCP.
The French company check is limited to French companies and is not a complete KYC or KYB compliance check. The described scope does not establish global company coverage, sanctions screening, or beneficial-ownership checks. Do not present a result from this tool as a substitute for those checks.
A receipt can help verify integrity, not truth
A receipt ID can help retrieve the same envelope and check, using the service’s content-addressed mechanism, whether the returned envelope has changed. That is an integrity and retrieval aid. It does not prove that the underlying source facts were accurate, certify a claim as true, or amount to a digital signature.
Rank #3
- Product type: Screw kit
- Made by Super Micro
- Manufacturer part number: MCP-410-00005-0N
- Supermicro MCP-410-00005-0N Screw Bag(100PCS) and Label for 24x Hot swap
- Mfr Part Number: MCP-410-00005-0N
Keep those questions separate in the agent’s response: “Is this the same envelope I retrieved earlier?” is not the same as “Was the source correct?” The evidence source and its quality still matter.
Free tools Windows power users keep installed
One-click scans. No signup required.
Where MCP ends and the envelope begins
MCP defines the client-server protocol and transports. It does not prescribe FACTRAIL’s evidence fields or require applications to use evidence envelopes. The envelope is the server’s application-level response design, carried through MCP.
The official MCP transport documentation describes Streamable HTTP as a way to carry protocol messages between client and server. In the 2026-07-28 documentation, messages are POSTs to a single MCP endpoint, and replies may be JSON or request-scoped SSE. It also requires servers to validate the Origin header to help prevent DNS rebinding. These details are tied to that documented revision; check the relevant specification revision rather than treating transport behavior as immutable.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Service figures are time-sensitive
The FACTRAIL tutorial reports seven tools in the endpoint’s listing at the time checked, while detailing four tools. It also reports a server response using version 2.4.1 and an example envelope captured on 2026-09-29; the tutorial recounts an earlier capture using version 2.4.0. These are dated service snapshots, not lasting MCP facts.
For the period the tutorial identifies as 2026-09-29, it reports a limit of 60 MCP requests per minute per client IP and a separate 600 requests per minute for traffic through known MCP gateways. The tutorial also lists Apify Actor wrapper prices of $0.005 per successful verification and $0.03 per successful import assessment, checked on that date. Those prices apply to the described third-party wrapper, not the direct endpoint. Verify limits and prices against the service before relying on them; either may change.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
