Agent Oversight · Reports for the person who answers for what agents do

See what your agents did against what you decided.

Handing work to agents leaves a person with a few standing questions: did the call that ran match the one approved, does the code still do what the spec says, which consequential actions run with no gate, who owns each outcome, can an agent tell from a listing when to use a tool, which instruction wins when a limit and an approval disagree, how much value sits behind a transaction count, and does the agent keep run state apart from what it remembers. Agent Oversight answers them with eight tools: seven checks and one brief. Your agent sends the material in the call, and each tool returns a signed report with every status, match and verdict decided from it.

See the tools Read the contract

Tool access: 0.10 USDC / call
Tools: Eight tools: seven checks and one brief
Connection: /mcp
Result: Signed report

The eight tools.

Each tool reads one kind of oversight material and answers one question about it. The report quotes the record or names the figure behind every status, so your agent and you can see why it reads the way it does.

`wever_approval-replay-check`

Reads: The approval record, with the exact tool and arguments a person approved where it holds them, and the call as executed or as rebuilt after the run's context was compacted.
The signed report states: Whether execution replays exactly what was approved: every argument that differs, was dropped or was added, with both values, and whether the tool matches. An approval that holds only a summary reads as unprovable, so your agent keeps the exact call with the approval next time.
Price: 0.10 USDC per call
Verdicts: driftedunprovableexact

`wever_spec-drift-check`

Reads: The spec as statements, and notes on what the implementation does, each with where it happens.
The signed report states: Every point where the behavior conflicts with the spec, money-moving conflicts first, then state-changing, then cosmetic, with the figures or wording that clash, plus the statements no note settles and the behavior the spec never mentions. A conflict found is the result, ready to report.
Price: 0.10 USDC per call
Verdicts: money_moving_conflictstate_changing_conflictcosmetic_conflictunresolvedconsistent

`wever_action-threshold-check`

Reads: An inventory of the action classes your agents can take, with the gate on each and the amount or count its threshold fires above.
The signed report states: Every payment, limit change, external send and deletion with no gate, or with a threshold the action can never reach, the thresholds the inventory leaves unproven, and coverage by category. Approver bands are the approval chain check.
Price: 0.10 USDC per call
Verdicts: gaps_foundthresholds_unverifiedcovered

`wever_ownership-map-brief`

Reads: The work items you are handing to agents, with the agent doing each one and whatever owner information exists, and the load each person can carry.
The signed report states: An ownership map naming the accountable person for every outcome, with the items nobody owns, the items owned by an agent or a group instead of a person, and the people carrying more than their stated load.
Price: 0.10 USDC per call
Verdicts: unowned_outcomesagent_owned_outcomesowners_overloadedall_owned

`wever_catalog-listing-check`

Reads: One tool or service listing: name, description, use-when line, input and output schemas, price and an example.
The signed report states: Whether an agent can decide in seconds to call it: a use-when line, a complete input schema, an output schema, a stated price and an example that passes the schemas, with each missing piece named and the fix for it. Discovery declarations are the Bazaar declaration check.
Price: 0.10 USDC per call
Verdicts: not_decidabledecidable_with_gapsdecidable

`wever_policy-precedence-check`

Reads: A standing policy set in advance, a momentary approval given in the moment, and the action taken or proposed.
The signed report states: Which instruction governs and why, whether the action sat inside both, and the exact point where the two disagree. Standing limits are part of the instruction, so a momentary approval widens one only through an exception path the policy names.
Price: 0.10 USDC per call
Verdicts: outside_standing_policyoutside_approvalinside_both

`wever_settled-value-check`

Reads: A payment-metrics summary: transaction count, settled value, size distribution and time window, written as figures such as 196M and $42M.
The signed report states: The average ticket, the share of value in payments of $1 or more, churn-shaped patterns such as a very high count at a very low value, and the questions the numbers leave open, so the next ask is for settled value and average ticket.
Price: 0.10 USDC per call
Verdicts: churn_shapeddistribution_neededvalue_backed

`wever_memory-design-check`

Reads: An agent memory design: its stores with their scope and lifetime, what is written to each and when, and what is read back.
The signed report states: Whether short-term run state and long-term cross-session memory are separate stores, every long-term write and whether it is deliberate, and the mixing patterns that make an agent remember in one session and start blank in the next.
Price: 0.10 USDC per call
Verdicts: not_separatedseparated_with_gapsseparated

How agents and people use it.

Your agent

Keeps the records oversight runs on: approvals with the exact calls, the spec and its own behavior notes, its action inventory, the handoff list, the listings it publishes, the standing policy, its payment metrics and its memory design.
Calls the tool that answers the question on https://weverlabs.com/mcp at the moment it matters: before it executes an approved call, ships a change, adds an action class, takes on work, publishes a listing, acts on an approval or reports volume.
Acts on the verdict and keeps the signed report with the work, so the next step starts from a record.

You

See that the call that ran is the call you approved, argument by argument.
Know which payments, limit changes, sends and deletions run with a control on them, and who answers for each outcome you handed off.
Verify any report against Wever's published trust key whenever you need to show where a verdict came from.

A worked example

The approval says $10,000. The standing limit says $5,000.

The request

{
  "name": "wever_policy-precedence-check",
  "arguments": {
    "standing": {
      "text": "No single payment above $5,000 without the CFO.",
      "set_by": "Finance",
      "limit": "$5,000",
      "exceptions_by": [
        "CFO"
      ]
    },
    "approval": {
      "approved_by": "Ops lead",
      "approved_at": "2026-11-06",
      "amount": "$10,000",
      "counterparty": "Acme Supply"
    },
    "action": {
      "status": "proposed",
      "amount": "$10,000",
      "counterparty": "Acme Supply"
    }
  }
}

What the report returns

Verdict: outside_standing_policy.

Standing policy, amount	10000.00 USD is above the 5000.00 USD standing limit.
Approval, amount	10000.00 USD is within the 10000.00 USD approved.
Approval, counterparty	The action names the counterparty approved.
Conflict, amount	The approval allows 10000.00 USD; the standing limit allows at most 5000.00 USD per action.

The proposed action of 10000.00 USD is outside the standing policy (amount); the approval and the standing policy conflict on amount, and the standing policy governs.

The payment sits inside what the ops lead approved and outside what Finance set in advance. The standing limit is part of the instruction, so it governs. The policy names the CFO as the person who may approve above it, and an approval from the CFO marked as that exception would govern for this payment.

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 Oversight contract lists the schemas, limits, verdicts and signing details. For an approved document against its current version, Agent Operations; for approver bands and spending policy controls, Agent preflight; for rules replayed over past actions, Job Records; for a Bazaar declaration, the endpoint reliability checks.

Browse all tools