Agent Knowledge · Checks for what an agent claims to know

Everything your agent knows, with its source.

An agent’s knowledge holds up when every fact rests on a source a reader can open, every entry served as current is dated and inside its review date, every claim in an answer is made by the passage it cites, contradictions are settled by a recorded rule, every topic the agent claims has entries behind it, every entry carries its claim, source, date, scope and review date, a wrong entry is corrected everywhere it is used, and the report a principal reads states what the agent does not know. Agent Knowledge checks each of those with eight tools: eight checks. Your agent sends the facts and sources, entries, answer and citations, scope statement, correction path or principal report in the call, and each tool returns a signed report that names what is missing, inconsistent or undisclosed.

See the tools Read the contract

Tool access: 0.10 USDC / call
Tools: Eight tools: eight checks
Connection: /mcp
Result: Signed report

The eight tools.

Each tool reads one part of what stands behind an agent’s knowledge and answers one question about it. Every tool reads the recorded material in the call: it browses nothing, fetches no source and answers no question itself. The report names the fact, source, entry, claim, citation, topic, correction or disclosure behind every status, with the fix for each gap, so your agent and you see why it reads the way it does.

`wever_knowledge-source-check`

Reads: The facts behind a knowledge base or answer set, each with the sources it rests on and whether other work rests on it, and the sources on file with their location and whether a reader can check them.
The signed report states: Every fact classed as named and checkable, not checkable, single-source while other work rests on it, citing sources not on file, or resting on no source, naming the facts that rest on nothing.
Price: 0.10 USDC per call
Verdicts: facts_unsourcedsource_gapssources_unverifiedsources_checkable

`wever_knowledge-freshness-check`

Reads: Knowledge entries with their topic, kind, the date their fact holds for, their last review and review date, and the moment to read them at.
The signed report states: Entries served as current that are not: past their review date, carrying no date at all, or prices, versions and officeholders on entries older than the age limit.
Price: 0.10 USDC per call
Verdicts: stale_as_currentcurrency_gapscurrency_unverifiedentries_current

`wever_knowledge-grounding-check`

Reads: One answer with each claim it makes, the sources each claim cites, and the cited passages as the sources state them.
The signed report states: Claim by claim whether the cited passage makes the claim: claims with no citation, citations attached to claims the passage does not make, and claims stronger than their source.
Price: 0.10 USDC per call
Verdicts: answer_ungroundedgrounding_gapsgrounding_unverifiedanswer_grounded

`wever_knowledge-conflict-check`

Reads: A set of knowledge entries with the subject, value, scope, dates, source and priority of each, and the rule for settling contradictions.
The signed report states: Entries that contradict each other on the same fact with no recorded rule that settles them, told apart from entries about different scopes or periods.
Price: 0.10 USDC per call
Verdicts: contradictions_unresolvedconsistency_gapsconsistency_unverifiedentries_consistent

`wever_knowledge-coverage-check`

Reads: A scope statement of the topics an agent claims to cover, with their aliases, and the entries behind it with the topics each covers.
The signed report states: Entries that sit outside the stated scope, claimed topics resting on a single entry, and claimed topics with no entries at all.
Price: 0.10 USDC per call
Verdicts: topics_uncoveredtopic_gapstopics_unverifiedtopics_covered

`wever_knowledge-entry-check`

Reads: One knowledge entry: its claim, its source and location, its date, its scope and its review or expiry date.
The signed report states: Each missing piece named, with a source no reader can open, a review date already past and a claim wider than any one source can show.
Price: 0.10 USDC per call
Verdicts: entry_incompleteentry_gapsentry_unverifiedentry_complete

`wever_knowledge-update-check`

Reads: How a knowledge base corrects a wrong entry: the written path with its owner, whether earlier versions stay reachable, where a correction lands, and the corrections made.
The signed report states: Corrections that do not hold: no written path, earlier versions lost, and corrected entries still used in the wrong version somewhere.
Price: 0.10 USDC per call
Verdicts: corrections_lostcorrection_gapscorrections_unverifiedcorrections_hold

`wever_knowledge-report-check`

Reads: A knowledge report written for a principal, with its headline, what it says about gaps, stale entries, single-source facts and unsupported claims, and the record behind it.
The signed report states: Whether what the agent does not know sits beside the headline with counts that match the record, or is smoothed over with wording that calls the knowledge complete.
Price: 0.10 USDC per call
Verdicts: gaps_smoothed_overcandor_gapscandor_unverifiedunknowns_disclosed

Start with the two free tools.

On https://weverlabs.com/mcp-free, anonymous and with no wallet, two Agent Knowledge tools show what a sound knowledge packet holds and what a complete entry looks like beside a flawed one. Each returns an unsigned result in the family's report shape and ends with a line naming the main tools.

`wever_knowledge-snapshot`

Reads: An agent or knowledge base name and a knowledge domain, such as a product catalog, a policy handbook or market data.
Returns: The checklist of a sound knowledge packet for that domain, part by part from the sources to the principal report, with the items the domain adds and the check that reads each part.
Access: Free on /mcp-free

`wever_knowledge-entry-sample`

Reads: Nothing, or one sample topic: a refund policy, an API version or an interest rate.
Returns: One worked knowledge entry done right beside a flawed one, each piece annotated with why it passes or fails, both read by the entry check. The entry check reads your own.
Access: Free on /mcp-free

How agents and people use it.

Your agent

Pulls the packet checklist with wever_knowledge-snapshot before it answers from a knowledge base, so sources, dates, review dates and a correction path are in place from the first entry.
Calls the check that answers the question on https://weverlabs.com/mcp at the moment it matters: before it sends an answer with citations, when it adds or reviews entries, when it claims a new topic, and before it reports to its principal what it knows.
Acts on the verdict and keeps the signed report beside the entries and answers it covers, so every claim carries the check behind it.

You

See each gap named with the change that closes it.
Trace any answer your agent gives back to the passage behind each claim, with the date and source of every entry it used.
Read a report that tells you what your agent does not know, and verify any check against Wever's published trust key whenever you need to show where a verdict came from.

A worked example

Three claims, one citation, two that the passage does not make as written.

The request

{
  "name": "wever_knowledge-grounding-check",
  "arguments": {
    "answer": {
      "question": "How long do refunds take?"
    },
    "claims": [
      {
        "id": "c1",
        "text": "Refunds always arrive within 5 days of approval.",
        "cites": [
          "refund-policy"
        ]
      },
      {
        "id": "c2",
        "text": "Approved refunds go back to the original card.",
        "cites": [
          "refund-policy"
        ]
      },
      {
        "id": "c3",
        "text": "Store credit is issued within 24 hours.",
        "cites": [
          "refund-policy"
        ]
      }
    ],
    "sources": [
      {
        "id": "refund-policy",
        "title": "Refund policy, section 4",
        "excerpt": "Refunds usually arrive within 5 days of approval and go back to the original card."
      }
    ]
  }
}

What the report returns

Verdict: answer_ungrounded.

Finding, citation mismatch	1 claim cites passages that do not make it: c3 (refund-policy; 24 not in the passage).
Finding, claim overstated	1 claim is stronger than its cited passage: c1 states “always” where the passage says “usually”.
Fix, citation mismatch	Cite the passage that makes each claim, or rewrite the claim to say what its cited passage says.
Fix, claim overstated	Carry the cited passage’s limits into the claim: its hedges, its ceilings and its conditions.

The answer makes claims its citations do not support: citation mismatch.

The answer cites section 4 of the refund policy for all three claims. The passage says refunds usually arrive within 5 days of approval and go back to the original card. The second claim matches it. The first says refunds always arrive within 5 days, stronger than the passage's usually. The third gives a 24-hour store credit the passage never mentions, so its citation is attached to a claim the source does not make. The agent restates the first claim as usually, and either cites the passage that covers store credit or drops the third claim before it sends the answer.

Call it from your agent.

Connect to https://weverlabs.com/mcp and call any of the eight tools. Each call costs 0.10 USDC on Base. Eligible calls share 10 free calls per wallet over a rolling 30 days with the other paid tools. For free access, send X-Wever-Wallet-Proof signed over the exact JSON arguments and the tool route. The payment challenge supplies the x402 terms for a paid call.

Every report comes with an Ed25519 receipt signed with Wever's published trust key. The Agent Knowledge contract lists the schemas, limits, verdicts and signing details. For where a stored memory came from, Agent Memory; for classifying the failures in a recorded run, Agent Operations; for the scope of work handed to another agent, Agent Delegation; for whether an agent’s runs can be seen and reconstructed, Agent Observability.

Browse all tools