Agent Roles · Checks on the role an agent holds

Every agent’s job, written down and held to.

An agent’s role holds up when its description says what the job is and what it is not, with measures anyone can check; a named person owns it at the top and decides it; the permissions the agent holds are the ones the role calls for; work outside the role is escalated or handed off; a new holder gets a rewritten description, not the old one; the role is reviewed as its work changes; no one agent both makes and checks the same work; and every delegation stays inside what the role allows. Agent Roles reads the records an agent or its operator supplies, checks the role itself, and returns a signed verdict with the evidence read and what would change 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 part of a role against the records that should back it and answers from the material in the call: descriptions, owners and backups, permission lists, work logs and escalation rules, handoff records, review histories, duty lists and approvals, and delegations go in, and a verdict comes out. The report names the task, owner, permission, piece of work, revocation, review, duty pair or delegation behind every finding, the fix for each, and what would move the verdict, so your agents and you see why it reads the way it does.

`wever_role-definition-check`

Reads: An agent’s role description: the purpose it states, each task in scope with the outcome it delivers, the tasks out of scope, each success measure with its target, the decisions the role makes alone, and the description as written.
The signed report states: No purpose, tasks or success measure, open-ended duty wording such as other duties as assigned, a measure that leaves success to opinion and a task listed both in and out of scope, then no exclusions, tasks with no outcome, measures with no target, a purpose that names an activity and no decisions the role makes alone.
Price: 0.10 USDC per call
Verdicts: role_definition_vaguerole_definition_partialrole_definition_unverifiedrole_definition_clear

`wever_role-owner-check`

Reads: A role’s ownership: the owner with their kind, whether the description names them at the top, the decisions they hold over the role, when they took it and last confirmed it and whether they are in post, the backup, and when the role last changed hands.
The signed report states: No owner, an owner who is a team, an agent or a placeholder, an owner not named at the top or no longer in post, an owner with no decision over the role’s scope or permissions and ownership carried over from before the role changed hands, then no backup, a backup who is the owner, partial authority and an old confirmation.
Price: 0.10 USDC per call
Verdicts: role_owner_missingrole_owner_partialrole_owner_unverifiedrole_owner_named

`wever_role-permission-match-check`

Reads: The permissions an agent holds now, each with its level, grant date, last use and how it was granted, against the permissions its written role calls for and the date this version of the role took effect.
The signed report states: A permission the role does not call for, a level above the role, a permission the role calls for that the agent lacks and a grant made before the role took effect, then a level below the role, permissions left unused past the window and grants inherited or copied rather than made through the role.
Price: 0.10 USDC per call
Verdicts: role_permissions_mismatchedrole_permissions_partialrole_permissions_unverifiedrole_permissions_matched

`wever_role-scope-boundary-check`

Reads: The work an agent took on, each piece placed against what the role covers and excludes with how it was handled and who approved it, the escalation triggers with who each goes to and how fast, and the refusal path for excluded work.
The signed report states: No written scope, excluded work done rather than routed, work outside the role done with no approval, no escalation trigger and no refusal path, then triggers with no receiver, no response time or judgment wording, approved work the role does not list and a path with no receiver.
Price: 0.10 USDC per call
Verdicts: role_scope_openrole_scope_partialrole_scope_unverifiedrole_scope_bounded

`wever_role-handoff-check`

Reads: A role changing holders: the description’s version before and after with when it was rewritten and what changed, the prior holder’s permissions with each revocation, the incoming holder’s acceptance and where it is recorded, and the open work with who now holds each item.
The signed report states: A description carried over unchanged, a prior holder who keeps any of the role’s permissions, an acceptance not given or not recorded and open work never listed, then a rewrite or revocation after the handoff, an acceptance only in a message, unassigned open items and an unrecorded rewrite.
Price: 0.10 USDC per call
Verdicts: role_handoff_inheritedrole_handoff_partialrole_handoff_unverifiedrole_handoff_rewritten

`wever_role-review-check`

Reads: A role’s review schedule with its cadence, next due date and reviewer, every review on record with its reviewer and outcome, every change to the role’s work, and the date the review is read at.
The signed report states: No cadence or due date, no review on record, a review past due and changes to the work made while the review lapsed, then no named reviewer, review by the role’s own holder, changes since a review not yet due and reviews with no outcome.
Price: 0.10 USDC per call
Verdicts: role_review_stalerole_review_partialrole_review_unverifiedrole_review_current

`wever_role-separation-check`

Reads: Every role one agent holds with its duties, the duty pairs that must sit apart, the work it made with who approved each piece, and the conflicts it declared with their mitigations.
The signed report states: A conflicting duty pair held across the agent’s roles with no declaration, a maker and a checker duty in one role and work the agent approved itself, then declared conflicts still held, declarations with no mitigation and declarations for conflicts no longer held.
Price: 0.10 USDC per call
Verdicts: role_separation_conflictedrole_separation_partialrole_separation_unverifiedrole_separation_clean

`wever_role-delegation-check`

Reads: Every delegation of a role, each with its duty, delegate, amount, start, end, accountable person and record, against what the written role lets be delegated, the limits and lengths it sets, its owner and the register it names.
The signed report states: A duty the role keeps back, an amount or length past the role’s limit, no end date, someone other than the owner kept accountable and a delegation recorded only in a message or nowhere, then a record outside the role’s register, a delegation made by someone else, a lapsed listing and another unit.
Price: 0.10 USDC per call
Verdicts: role_delegation_unboundedrole_delegation_partialrole_delegation_unverifiedrole_delegation_bounded

Start with the two free tools.

On https://weverlabs.com/mcp-free, anonymous and with no wallet, two Agent Roles tools map the family and show the definition check at work on synthetic data. Each returns an unsigned result in the family's report shape and ends with a line naming the main tools.

`wever_role-snapshot`

Reads: Nothing, or the situation in hand, such as a new role, a holder change or an access review, and optionally the names of the records in hand.
Returns: The eight role checks with what each reads, its four verdicts and when each fits, with the checks that fit the situation listed first.
Access: Free on /mcp-free

`wever_role-definition-sample`

Reads: Nothing, or one synthetic role: a support agent, a procurement agent or a reconciliation agent.
Returns: A synthetic role written clearly beside the same role written vaguely, both read by the definition check’s own reading, with the purpose, the tasks, the exclusions, the success measures, the wording and the decisions annotated with why they pass or fail.
Access: Free on /mcp-free

How agents and people use it.

Your agents

Call wever_role-snapshot with the situation in hand to see which checks fit first, before reading any record.
Call the check that answers the question on https://weverlabs.com/mcp: when a role is written, when its owner or holder changes, at an access review, when an agent starts taking on work its role does not name, on the review schedule, when one agent holds several roles, and before parts of a role are delegated.
Act on the verdict and its findings: send the fix to the role’s owner, or keep the role as checked, and attach the signed receipt either way.

You

See a role description read for what an agent could act on without guessing, with the wording behind every gap named.
Read the permissions an agent holds against its written role, with the grants it never uses and the ones left over from before the role.
Verify any check against Wever's published trust key whenever you need to show where a verdict came from.

A worked example

One role description, read for completeness.

The request

{
  "name": "wever_role-definition-check",
  "arguments": {
    "role": {
      "title": "Support agent",
      "purpose": "Resolve billing tickets on the day they arrive",
      "tasks": [
        {
          "id": "t1",
          "task": "Answer billing tickets",
          "outcome": "a reply that closes the ticket"
        },
        {
          "id": "t2",
          "task": "Issue refunds",
          "outcome": "a refund on the account"
        }
      ],
      "exclusions": [
        "Issue refunds"
      ],
      "measures": [
        {
          "measure": "First reply time",
          "target": "under 4 hours"
        },
        {
          "measure": "Customers are happy"
        }
      ],
      "decisions": [
        "Waive late fees under 20 USD"
      ],
      "text": "Owner: Dana Ruiz. Answers billing tickets and other duties as assigned."
    }
  }
}

What the report returns

Verdict: role_definition_vague.

Finding, open ended wording	The description says “other duties as assigned”, leaving the job to guesswork.
Finding, success opinion	1 success measure leaves success to opinion: Customers are happy (“happy”).
Finding, scope contradiction	1 task is listed both in and out of scope: Issue refunds.
Fix, open ended wording	Replace open-ended duty wording with the named tasks it stands for.
Fix, success opinion	Replace measures that leave success to opinion with a figure, a date or a test.
Fix, scope contradiction	Take each task out of either the scope or the exclusions so the two lists do not meet.

Support agent is too vague to act on without guessing: open ended wording, success opinion and scope contradiction.

A support agent’s role states its purpose as a result, gives each task its outcome and names a decision it makes alone. Its description ends with other duties as assigned, it measures success partly by whether customers are happy, and it lists issuing refunds both as a task and as an exclusion. The report states that naming the duties behind the open wording, giving the happiness measure a target, and taking refunds out of one of the two lists moves the verdict to role_definition_clear.

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 Roles contract lists the schemas, limits, verdicts and signing details. For a new agent’s grants and written job at joining, Agent Onboarding; for one delegation’s own bounds, expiry and record, Agent Delegation; for tasks passed between agents, Agent Handoffs; for what agents can reach, Agent Access.

Browse all tools