Agent Governance · Checks for the conditions agents run under

Set the terms before agents act.

Agents that buy, approve and run in production need their conditions set in advance: a sandbox that stays apart from production, caps with a scope and a breach rule, an offer every buying agent can find, liability agreed before the purchase, an approval record the executor runs exactly, a budget with all five controls, a named owner for every production agent, and purchase terms that match the contract. Agent Governance checks each with eight tools: eight checks. Your agent sends the design, setup, terms or records 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: eight checks
Connection: /mcp
Result: Signed report

The eight tools.

Each tool reads one kind of governance material and answers one question about it. The report names the credential, cap, surface, term, step, control, agent or purchase behind every status, so your agent and you can see why it reads the way it does.

`wever_sandbox-isolation-check`

Reads: A test or sandbox environment for agents: the credentials it holds and where each is accepted, the systems and tool platforms the agents use with what each accepts and its rate limit, the data the sandbox can read and its outbound default.
The signed report states: Every path from the test environment into production systems or data, production asset by production asset: credentials accepted in production, predictable or shared credentials, systems that take a sandbox credential or none, production data in reach, open outbound reach and systems with no rate limit, with the fix for each.
Price: 0.10 USDC per call
Verdicts: production_reachableisolation_unverifiedisolated

`wever_exposure-cap-check`

Reads: An exposure or spend cap design: each cap's amount, what it totals over (one order, one agent, one counterparty or the portfolio), its window and what happens to a request above it, plus requests that met a cap.
The signed report states: Every cap with no stated scope or window, every breach left undefined or allowed, and every downsized, split or queued request with no written rule or no record of the change.
Price: 0.10 USDC per call
Verdicts: cap_undefinedbreach_unruledcaps_unverifiedcaps_defined

`wever_discovery-coverage-check`

Reads: A seller's agent-discovery setup: the offer with its price and terms, and every registry, directory, agent card, feed and server where buying agents find it, with what each shows and how changes reach it.
The signed report states: Every surface where the offer is missing, stale or contradicts the offer or another surface on price or terms, and every single surface whose failure makes the seller invisible to buying agents.
Price: 0.10 USDC per call
Verdicts: contradictorycoverage_gapscoverage_unverifiedconsistent

`wever_liability-terms-check`

Reads: The terms between an agent's operator and a seller or platform: when liability for agent error is agreed, who pays and per error case, the cap, the dispute path, what the payment record proves, and every way a purchase can proceed.
The signed report states: Every purchase that can proceed with no agreed owner of agent error, liability left to be read from a protocol log after the fact, and every missing cap or dispute path.
Price: 0.10 USDC per call
Verdicts: liability_unagreedliability_gapsliability_unverifiedagreed_before_purchase

`wever_approval-record-check`

Reads: An approval-and-execution design: what the approval record holds and where it lives, each step that carries the call to the executor, and what the executor runs and checks.
The signed report states: Approvals recorded as summaries or kept in conversation history, execution that rebuilds the call instead of running the recorded one, and every point where the executed call can differ from the approved call without detection.
Price: 0.10 USDC per call
Verdicts: divergence_undetectedfidelity_gapsfidelity_unverifiedexecutes_as_approved

`wever_budget-controls-check`

Reads: An agent budget setup: its maximum per payment, its session or daily budget, its permitted recipients, its revocation path and its spending record.
The signed report states: Each of five controls present, weak or missing, with budgets that cap totals and leave recipients open, revocation that waits, and spending records nobody can read while money is moving.
Price: 0.10 USDC per call
Verdicts: controls_missingcontrols_weakcoverage_unverifiedfull_coverage

`wever_incident-ownership-check`

Reads: An agent operations setup: each agent with its named owner, who responds and who is paged when it errs, and the fallback that takes over when it is wrong or down.
The signed report states: Every production agent with no named person as owner, every error path with no responder or no one paged, and every agent with no fallback, or one nobody runs or has exercised.
Price: 0.10 USDC per call
Verdicts: unowned_agentsresponse_gapsownership_unverifiedowned_before_incident

`wever_terms-match-check`

Reads: Purchases with their quoted or recorded terms beside the governing published or contract terms: price, scope and cancellation.
The signed report states: Every mismatch between the terms a purchase was made on and the governing terms, with the price difference in percent, and every purchase paid with no comparison before payment.
Price: 0.10 USDC per call
Verdicts: terms_mismatchcompared_latematch_unverifiedterms_match

How agents and people use it.

Your agent

Keeps the material each check reads: its sandbox design, its caps, its discovery surfaces, its terms with each seller or platform, its approval and execution design, its budget setup, its operations setup and the terms of each purchase.
Calls the check that answers the question on https://weverlabs.com/mcp at the moment it matters: before a benchmark run, when a cap changes, when an offer changes, before the first purchase with a new counterparty, when the approval path changes, when a budget is issued, before an agent goes to production and before a purchase is paid.
Acts on the verdict and keeps the signed report with the work, so the next step starts from a record.

You

See each gap named with the change that closes it.
Know which paths lead from the sandbox to production, which caps hold, who owns each production agent and which purchases match their contract.
Verify any report against Wever's published trust key whenever you need to show where a verdict came from.

A worked example

A cap that downsizes. A cap that never resets.

The request

{
  "name": "wever_exposure-cap-check",
  "arguments": {
    "caps": [
      {
        "name": "Per-order cap",
        "amount": "40,000 USD",
        "scope": "order",
        "on_breach": "downsize",
        "records_change": false
      },
      {
        "name": "Desk exposure cap",
        "amount": "250,000 USD",
        "scope": "agent",
        "on_breach": "reject"
      }
    ],
    "cases": [
      {
        "cap": "Per-order cap",
        "requested": "48,600 USD",
        "executed": "24,300 USD",
        "outcome": "downsized",
        "recorded": false
      }
    ]
  }
}

What the report returns

Verdict: cap_undefined.

Scope, no window	Desk exposure cap totals per agent with no stated window, so when the total resets is undefined.
Breach, unwritten rule	Per-order cap downsizes a request above it with no written rule, so the size that goes through is decided at the moment.
Breach, unrecorded change	Per-order cap changes a request above it and records no change, so the original request and the reason leave no trace.
Fix, Per-order cap	Write the rule Per-order cap applies to a request above it, such as the size it downsizes to.
Fix, Per-order cap	Record every request Per-order cap changes, with its original size, its new size and the reason.
Fix, Desk exposure cap	State the window Desk exposure cap totals over, such as 24 hours, or that it never resets.

1 scope gap leaves a cap undefined: no window, with 4 breach gaps.

An agent places a 48,600 USD order against a 40,000 USD cap per order. The cap halves it to 24,300 USD, and no written rule says why half, and nothing records the original order. A second cap limits the desk to 250,000 USD per agent with no window, so no one can say when the total resets. Writing the downsizing rule, recording every changed order and stating the window gives both caps a defined scope and a ruled breach.

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 Governance contract lists the schemas, limits, verdicts and signing details. For refund rules, approval conflicts and gateway coverage, Agent Accountability; for sandbox escape routes, firewall order and approval durability, Agent Control; for approval replay and ownership maps, Agent Oversight.

Browse all tools