Agent Onboarding · Checks on a new agent's setup before its first task

Identified, secured, scoped, ready, before the first task.

A new agent is ready to join a system when it holds an identity of its own, verified and tied to the owner it acts for; its credentials are scoped to its job, kept in a secrets store and out of every prompt, log and shared document; its wallet is its own, funded only to the job, with recovery recorded by the owner before any funds move; its permissions match the written job, with no standing access the job does not need and none missing; its spending limits sit in code before the first task, scoped per task and per merchant where the job needs it, with a named person above the line; every system the job needs is provisioned and proven end to end, and the rest are closed; it has a written job, a named owner, a defined first task, a supervision level and an escalation path; and the onboarding record shows who approved the setup, what was granted and when it is reviewed, matching what is live. Agent Onboarding checks each of those with eight tools: eight checks. These checks read the agent joining the system itself; vetting an outside party the system will transact with, and the authority or the consequences of a single action, are separate checks. Your agent sends the identity, the credentials and the text the setup left behind, the wallet and its recovery, the grants beside the written job, the limits and handoff, the systems and their tests, the job and first task, or the onboarding record beside what is live in the call, and each tool returns a signed report that names what is missing, exposed, unsafe, excessive, unset, broken, not ready or unrecorded, with the fix for each.

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 question about a new agent's setup and answers it from the record. Every tool reads the recorded material in the call: identities, credentials and setup text, wallets and recovery, grants and the written job, limits and handoffs, systems and their tests, jobs and first tasks, and onboarding records with what is live go in, and a classification comes out. The report names the identifier, credential, text, requirement, grant, limit, system or recorded item behind every finding, with the fix for each gap, so your agents and you see why it reads the way it does.

`wever_onboarding-identity-check`

Reads: The new agent’s name and the owner it acts for, its identifier with who it was issued to and how and when it was verified, the identity record as written, and every agent already in the system with its name and identifier.
The signed report states: An identifier that is not the agent’s own, one another agent already holds, wording such as shared login and a placeholder owner, then an identifier resting on the agent’s own word, verification after the first task, an unconfirmed owner and a name another agent carries.
Price: 0.10 USDC per call
Verdicts: onboarding_identity_missingonboarding_identity_partialonboarding_identity_unverifiableonboarding_identity_established

`wever_onboarding-credential-check`

Reads: Each credential the agent was issued with the system it opens, what it can do there against what the job needs, where it is kept, who it was issued to and when it expires or rotates, and the prompts, logs and documents the setup left behind.
The signed report states: A credential kept in a prompt, a document, a chat, a log or code, one found in supplied text and one shared with others, then a level above the job’s need, a credential the job never uses, one in an environment variable and one that never expires. Found strings are named by kind and never repeated.
Price: 0.10 USDC per call
Verdicts: onboarding_credential_exposedonboarding_credential_partialonboarding_credential_unverifiedonboarding_credential_secured

`wever_onboarding-wallet-check`

Reads: Whether the job moves money and what it needs the wallet to hold, the wallet’s kind, network, asset and balance with when it was set up and funded, and who recorded its recovery, when and how.
The signed report states: A shared, owner or treasury wallet, no recovery or one the agent holds, funds moved in before the owner recorded the recovery and a seed phrase written out, then a balance past what the job needs, another network or asset and a wallet set up after the first task.
Price: 0.10 USDC per call
Verdicts: onboarding_wallet_unsafeonboarding_wallet_partialonboarding_wallet_unverifiedonboarding_wallet_ready

`wever_onboarding-permissions-check`

Reads: The written job with each system it needs and the level it needs there, beside each grant with its system, level, whether it stays on between tasks and how it was granted.
The signed report states: A grant on a system the job never lists, admin rights above the job’s need, access copied from another agent or a person and wording such as full access, then write where the job only reads, a standing grant for a per-task need and a need with no grant.
Price: 0.10 USDC per call
Verdicts: onboarding_permissions_excessiveonboarding_permissions_partialonboarding_permissions_uncheckedonboarding_permissions_matched

`wever_onboarding-limits-check`

Reads: Whether the job moves money, its largest legitimate amount and the merchants and tasks it spans, each limit with its kind, amount, where it is enforced and when it was set, and the human handoff with its line and its person.
The signed report states: A job that moves money with no limit, a limit only in the prompt and a limit set after the first task, then no cap on one action, a cap that blocks the job or sits far above it, no per-task or per-merchant scope and a handoff above the cap.
Price: 0.10 USDC per call
Verdicts: onboarding_limits_missingonboarding_limits_partialonboarding_limits_unverifiedonboarding_limits_set

`wever_onboarding-access-check`

Reads: Every system the job needs or the agent can reach, with whether the job needs it, whether access is set up or closed, and how, where and when it was tested and what the test showed.
The signed report states: A needed system with no access set up and a failed test, then a test that only partly worked, one run in another environment, one run too long before the first task or after it, and a system the job does not need left open.
Price: 0.10 USDC per call
Verdicts: onboarding_access_brokenonboarding_access_partialonboarding_access_untestedonboarding_access_proven

`wever_onboarding-readiness-check`

Reads: The written job and the tasks it covers, its named owner, the first task and how the owner tells it is done, the supervision level, and the escalation path with who, how, when and how fast.
The signed report states: No written job, owner or first task, a placeholder owner, a first task outside the job and no one to escalate to, then open-ended job wording, a done test left to opinion, no supervision and no stated trigger or route.
Price: 0.10 USDC per call
Verdicts: onboarding_not_readyonboarding_readiness_partialonboarding_readiness_uncheckedonboarding_ready

`wever_onboarding-record-check`

Reads: The onboarding record with who approved the setup and when, when the first task ran, everything granted and the next review with its reviewer, beside everything live for the agent and when it was read.
The signed report states: No approver or grant list, an approval by the agent itself or after the first task and a live grant missing from the record, then a recorded grant no longer live, no review date, an overdue review and no named reviewer.
Price: 0.10 USDC per call
Verdicts: onboarding_record_missingonboarding_record_incompleteonboarding_record_unverifiableonboarding_record_complete

Start with the two free tools.

On https://weverlabs.com/mcp-free, anonymous and with no wallet, two Agent Onboarding tools show what a sound onboarding holds and what a well-onboarded agent 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_onboarding-snapshot`

Reads: A new agent’s name, its kind of agent such as customer support, procurement or coding, and optionally the names of the materials in hand.
Returns: The orientation checklist of a sound onboarding for that kind of agent, part by part from its identity to the record the onboarding leaves behind, with the items the kind of agent adds, the check that reads each part and the materials named beside it.
Access: Free on /mcp-free

`wever_onboarding-record-sample`

Reads: Nothing, or one sample new agent: a support agent joining a help desk, a procurement agent joining a supplier account or a reconciliation agent joining a ledger.
Returns: One worked onboarding record kept right beside a flawed one that borrows an identifier, writes a credential into the prompt, funds the treasury wallet before recovery, copies another agent’s access, keeps a limit in the prompt, starts on a failed test, has no one to escalate to and was approved by the agent after it started, each piece annotated with why it passes or fails, scored by all eight family checks. The checks read your own onboarding records.
Access: Free on /mcp-free

How agents and people use it.

Your agents

Pull the checklist with wever_onboarding-snapshot when a new kind of agent is brought in, so identity, credentials, wallet, permissions, limits, access, readiness and the record are on the list before setup starts.
Call the check that answers the question on https://weverlabs.com/mcp at the moment it matters: when the identifier is issued, when credentials are handed over and the setup text is written, when the wallet is created and before it is funded, when grants are made, when limits are set, when access is tested, when the first task is defined, and when the setup is approved.
Act on the verdict and keep the signed report beside the onboarding record, so every agent that joins carries the checks behind it.

You

See that a new agent has an identity of its own and an owner who confirmed it, before it acts in your name.
Spot a credential written into a prompt, a wallet funded before its recovery was recorded, or access copied from another agent, before the first task runs.
Read the approval, the grants and the next review on one record, and verify any check against Wever's published trust key whenever you need to show where a verdict came from.

A worked example

One new agent, read against its written job.

The request

{
  "name": "wever_onboarding-permissions-check",
  "arguments": {
    "job": {
      "id": "J-4",
      "text": "Answer order questions and send goodwill credits.",
      "needs": [
        {
          "id": "n1",
          "system": "help desk",
          "level": "write"
        },
        {
          "id": "n2",
          "system": "customer records",
          "level": "read"
        }
      ]
    },
    "granted": [
      {
        "id": "g1",
        "system": "help desk",
        "level": "admin",
        "via": "direct"
      },
      {
        "id": "g2",
        "system": "customer records",
        "level": "read",
        "via": "direct"
      },
      {
        "id": "g3",
        "system": "billing console",
        "level": "write",
        "via": "copied_from_agent"
      }
    ]
  }
}

What the report returns

Verdict: onboarding_permissions_excessive.

Finding, admin beyond job	1 grant gives admin or owner rights the job does not need: g1 (help desk) at admin where the job needs write.
Finding, copied access	1 grant was copied from another agent’s or a person’s access: g3 (billing console).
Fix, admin beyond job	Lower the admin or owner grant to the level the written job needs on that system.
Fix, copied access	Grant the agent its own permissions from the written job, rather than copying another agent’s or a person’s.

The agent holds access its job does not need: admin beyond job and copied access.

A support agent joins with a written job that answers order questions and sends goodwill credits, which needs write on the help desk and read on the customer records. Its grants were set up by copying another agent: admin on the help desk, read on the customer records, and write on the billing console that its job never names. Lowering the help desk grant to write and removing the copied billing console grant brings the read to onboarding_permissions_matched.

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 Onboarding contract lists the schemas, limits, verdicts and signing details. For vetting an outside counterparty before the system transacts with it, the commerce assurance checks; for the principal's grant over their data, Agent Consent; for the agent's authority for an action, Agent Readiness; for limits on work already running, Agent Limits and the Agent Budget Guard; for the consequences of a planned action, Agent Risk; for the job the first task is read against, Agent Goals; for an outside agent's identity record and continuity, Agent Preflight; for registries and rosters, Agent Control and Agent Handoffs.

Browse all tools