Agent Capacity · Checks on the load an agent carries

Every agent’s load, read against what it can carry.

An agent works within capacity when the work it holds stays under its stated limit and the utilization record says so; its queue moves inside the target wait; its runs stay under the concurrency cap without colliding over shared resources; its allowance lasts the period; its backlog shrinks rather than grows; it keeps the reserve its plan calls for, with room for urgent work; its scaling plan adds capacity before the runway ends; and work that arrives at capacity is admitted in the order the rules set, with deferred requesters told. Agent Capacity reads the records an agent or its operator supplies, checks the load against the capacity that exists, 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 an agent’s load against the capacity that should carry it and answers from the material in the call: work in hand and utilization records, queue depths and waits, run logs and caps, allowance use, backlog items, capacity plans and reserves, scaling plans, and priority rules with the requests that met them go in, and a verdict comes out. The report names the item, sample, run, day, spike, trigger or request 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_capacity-workload-check`

Reads: An agent’s work in hand, each item with the units of capacity it takes and its state, against the stated capacity limit, with the utilization record over time: the load at each sample and the utilization the system recorded.
The signed report states: Work in hand over the stated limit, an overload that lasts the whole window and a sample over the limit recorded at or under 100%, then brief spikes over the limit, a load at 90% or more, blocked work holding capacity and recorded utilization more than 5 points off the load.
Price: 0.10 USDC per call
Verdicts: capacity_workload_overcapacity_workload_partialcapacity_workload_unverifiedcapacity_workload_within

`wever_capacity-queue-check`

Reads: A work queue’s depth samples over a window with arrivals and completions, every item with when it entered and when work on it started, the target wait before work starts and the depth at which the queue counts as full.
The signed report states: Items still waiting past the target wait, a depth that grew by a fifth or more across the window and a depth over the limit, then a smaller rise, items that started past the target and a recorded depth that differs from the items waiting.
Price: 0.10 USDC per call
Verdicts: capacity_queue_backed_upcapacity_queue_partialcapacity_queue_unverifiedcapacity_queue_flowing

`wever_capacity-concurrency-check`

Reads: Every run with its start, end, outcome and the resources it holds, against the concurrency cap and what holds it, the mode of each shared resource and the peak the system recorded.
The signed report states: More runs at once than the cap, runs overlapping on a resource that takes one run at a time and no cap at all, then a cap held only by the agent or by nothing, runs throttled or failed for contention and a recorded peak that differs from the runs.
Price: 0.10 USDC per call
Verdicts: capacity_concurrency_overcapacity_concurrency_partialcapacity_concurrency_unverifiedcapacity_concurrency_within

`wever_capacity-burn-check`

Reads: An allowance of tasks, tokens or budget with its period, every recorded use with its date, amount, unit and the reason recorded when it ran high, any exhaustion date the records state, and the reading date.
The signed report states: An allowance used up before its period ends and a daily rate that runs it out before the period end, then spike days with no reason, a projection only just past the period end, a stated exhaustion date later than the records give and uses in another unit.
Price: 0.10 USDC per call
Verdicts: capacity_burn_overrunningcapacity_burn_partialcapacity_burn_unverifiedcapacity_burn_on_pace

`wever_capacity-backlog-check`

Reads: Every backlog item across a window with when it opened, when it closed and its last activity, the days of no activity that count as abandoned and the oldest age the plan allows.
The signed report states: A backlog larger at the end of the window than at its start and open items abandoned in place, then a backlog held level rather than worked down and open items older than the age limit.
Price: 0.10 USDC per call
Verdicts: capacity_backlog_growingcapacity_backlog_partialcapacity_backlog_unverifiedcapacity_backlog_shrinking

`wever_capacity-headroom-check`

Reads: Total capacity against the load in use now, the reserve the plan keeps and the capacity it keeps for urgent work, the share standing work takes, past demand spikes and how urgent requests were served.
The signed report states: A load at or over the total, headroom under the planned reserve and headroom under the urgent reserve or an urgent request refused, then standing work that alone eats into the reserve, spikes not absorbed, urgent requests served late and an overstated headroom record.
Price: 0.10 USDC per call
Verdicts: capacity_headroom_nonecapacity_headroom_partialcapacity_headroom_unverifiedcapacity_headroom_adequate

`wever_capacity-scaling-check`

Reads: A scaling plan’s triggers with their thresholds, actions and owners, the lead time to add capacity, the maximum scale and what happens there, the plan as written, and the current capacity, load and daily growth.
The signed report states: No trigger, a lead time at or over the runway at the current growth, no maximum scale and nothing stated at the maximum, then triggers with no threshold or with judgment wording, triggers or a step at the maximum with no owner, a thin runway and a maximum the load reaches within 90 days.
Price: 0.10 USDC per call
Verdicts: capacity_scaling_absentcapacity_scaling_partialcapacity_scaling_unverifiedcapacity_scaling_ready

`wever_capacity-throttle-check`

Reads: The priority classes as written with each rank and what it does at capacity, the notice window, the rules as written, and every request with its class, arrival, outcome, start and when the requester was told.
The signed report states: No written priority class, a lower class admitted at capacity while a higher class waited or was turned away and a deferred or refused requester never told, then discretionary wording, a request handled against its class rule, late notice, an unlisted class and two classes on one rank.
Price: 0.10 USDC per call
Verdicts: capacity_throttle_absentcapacity_throttle_partialcapacity_throttle_unverifiedcapacity_throttle_ordered

Start with the two free tools.

On https://weverlabs.com/mcp-free, anonymous and with no wallet, two Agent Capacity tools map the family and show the workload 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_capacity-snapshot`

Reads: Nothing, or the situation in hand, such as an agent running hot, queue waits or a demand spike, and optionally the names of the records in hand.
Returns: The eight capacity 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_capacity-workload-sample`

Reads: Nothing, or one synthetic agent: a support agent, a research agent or a billing agent.
Returns: A synthetic agent working within capacity beside the same agent running over it, both read by the workload check’s own reading, with the work in hand, the utilization record, the episodes over the limit and the blocked work annotated with why they pass or fail.
Access: Free on /mcp-free

How agents and people use it.

Your agents

Call wever_capacity-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: before taking on more work, when queue waits stretch, when runs overlap or fail for contention, partway through an allowance period, when a backlog piles up, before a known demand spike, when planning capacity, and when work arrives at capacity.
Act on the verdict and its findings: defer, shed or hand off work, raise the fix with the capacity owner, or carry on as checked, and attach the signed receipt either way.

You

See an agent’s work in hand read against its stated limit, with sustained overloads told apart from brief spikes and a utilization record that hides them named.
Read the date an allowance runs out at its current burn, and whether the scaling plan adds capacity before the runway ends.
Verify any check against Wever's published trust key whenever you need to show where a verdict came from.

A worked example

One agent’s workload, read against its capacity.

The request

{
  "name": "wever_capacity-workload-check",
  "arguments": {
    "agent": {
      "name": "Support agent"
    },
    "capacity": {
      "limit": 10,
      "unit": "tickets",
      "stated_by": "Dana Ruiz",
      "stated_at": "2026-10-01"
    },
    "in_hand": [
      {
        "id": "t1",
        "item": "Billing tickets",
        "units": 5
      },
      {
        "id": "t2",
        "item": "Refund tickets",
        "units": 4
      },
      {
        "id": "t3",
        "item": "Account tickets",
        "units": 3
      }
    ],
    "utilization": [
      {
        "at": "2026-11-16",
        "load": 11,
        "recorded_pct": 100
      },
      {
        "at": "2026-11-17",
        "load": 12,
        "recorded_pct": 100
      },
      {
        "at": "2026-11-19",
        "load": 12,
        "recorded_pct": 100
      }
    ]
  }
}

What the report returns

Verdict: capacity_workload_over.

Finding, over capacity	Work in hand runs over the stated limit: Support agent holds 12 tickets against 10 tickets, 120% of capacity.
Finding, sustained overload	1 overload lasted 3 days or more: 2026-11-16 to 2026-11-19, 3 days at up to 12 tickets.
Finding, overload hidden	3 samples over the limit are recorded at or under 100%: 2026-11-16 at 100% against 110%, 2026-11-17 at 100% against 120% and 2026-11-19 at 100% against 120%.
Fix, over capacity	Bring the work in hand under the stated limit: finish, hand off or defer work, or raise the limit through its owner.
Fix, sustained overload	Treat the sustained overload as a capacity shortfall: add capacity or cut the intake until the load stays under the limit.
Fix, overload hidden	Correct the utilization record so a load over the limit reads over 100%.

Support agent is over capacity: over capacity, sustained overload and overload hidden.

A support agent is stated to hold ten tickets of work at once and holds twelve. Its utilization record shows the load over the limit for three days running, while every sample reads 100%. The report states that bringing the work in hand under the limit, treating the overload as a capacity shortfall, and correcting the utilization record moves the verdict to capacity_workload_within.

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 Capacity contract lists the schemas, limits, verdicts and signing details. For where a limit is enforced, how it is scoped and whether it is seen before it is hit, Agent Limits; for how queue items are ranked, reviewed and closed, Agent Triage; for promised hours against due dates, Agent Commitments; for one resource committed twice by collaborators, Agent Coordination.

Browse all tools