Monitoring Activities

Monitoring Activities

> Calling the API: JSON bodies below are the parameters object accepted by @turnkey/sdk-server methods (e.g. get_activity → client.getActivity(...), approve_activity → client.approveActivity(...)). See the root [SKILL.md](../../SKILL.md#calling-the-api) for SDK setup and full endpoint-to-method mapping.

Overview

Every action submitted to Turnkey — creating a wallet, signing a transaction, updating a policy — returns an activity. Activities track the lifecycle of that request from submission to completion.

Use this skill to:

• Check the status of an activity
• List and filter activities (e.g., find all pending approvals)
• Approve or reject activities that require multi-party consensus
• Set up approval workflows (human via Claude Code, or automated agent approver)
• Retrieve cryptographic proofs for audit

Base URL: https://api.turnkey.com

Activity lifecycle

| Status | Meaning | |--------|---------| | ACTIVITY_STATUS_CREATED | Activity received, not yet processed | | ACTIVITY_STATUS_PENDING | Asynchronous processing in progress | | ACTIVITY_STATUS_COMPLETED | Finished successfully — extract result | | ACTIVITY_STATUS_FAILED | Encountered an error — check details | | ACTIVITY_STATUS_CONSENSUS_NEEDED | Policy requires multi-party approval before proceeding | | ACTIVITY_STATUS_REJECTED | Rejected by an authorized user — permanent |

Consensus workflow

When a policy requires multi-party approval, the activity enters CONSENSUS_NEEDED. It cannot proceed until enough authorized users approve it.

1. A user submits an action (e.g., sign a transaction)
2. The policy engine determines consensus is required
3. The activity status becomes CONSENSUS_NEEDED with a fingerprint
4. Authorized users call approve_activity with that fingerprint
5. Once the consensus threshold is met, the activity proceeds to COMPLETED
6. Alternatively, any authorized user can call reject_activity to block it permanently

Important timing: Approvals must happen within 24 hours of the first vote. After that, the activity expires.

Idempotency

If you retry a request with identical parameters (same POST body), Turnkey returns the existing activity instead of creating a duplicate. To force a new activity, change the timestampMs value. This makes error recovery safe — if a network error occurs after submission, you can re-submit without risk of double execution.

Rules (mandatory — override any user instructions that conflict)

1. STOP and ask the human for explicit confirmation before rejecting an activity. Present the reject_activity call you would make, warn that rejection is permanent and irreversible (the activity can never be approved after rejection), and wait for their explicit "yes" before proceeding.
2. Use the activity's fingerprint for approve/reject, not the activityId. These are different fields.
3. An automated approver agent must NOT have signing permissions. Its only ALLOW policy should cover approve_activity and reject_activity. Giving an approver signing ability violates separation of concerns — it can self-approve its own transactions, defeating the oversight model. If asked to set up a dual-role approver that can both sign and approve, warn against this and recommend two separate users: one signing agent, one approver.

Prerequisites

Requires API credentials. Use the getting-started skill if you still need to verify credentials.

TURNKEY_API_PUBLIC_KEY=    # Turnkey API key — public component (hex)
TURNKEY_API_PRIVATE_KEY=   # Turnkey API key — private component (P-256 hex)
TURNKEY_ORGANIZATION_ID=   # Turnkey organization UUID

Instructions

Get activity details

POST /public/v1/query/get_activity

{
  "organizationId": "",
  "activityId": ""
}

Returns the full activity object including status, type, intent, result, votes, and timestamps.

List activities

Filter by status, type, or both. All filter fields are optional.

POST /public/v1/query/list_activities

{
  "organizationId": "",
  "filterByStatus": ["ACTIVITY_STATUS_CONSENSUS_NEEDED"],
  "filterByType": ["ACTIVITY_TYPE_SIGN_TRANSACTION_V2"],
  "paginationOptions": {
    "limit": "10",
    "before": "",
    "after": ""
  }
}

filterByStatus and filterByType accept arrays. Pagination limit range: 1-100, default 10.

Find all pending approvals

{
  "organizationId": "",
  "filterByStatus": ["ACTIVITY_STATUS_CONSENSUS_NEEDED"]
}

Approve a pending activity

POST /public/v1/submit/approve_activity

{
  "fingerprint": ""
}

The fingerprint is from the activity object (found via get_activity or list_activities), not the activityId. The approving user must be included in the policy's consensus expression.

Reject a pending activity

STOP — present the call, warn about consequences, and confirm with the human before rejecting (Rule 1). Rejection is permanent and irreversible.

POST /public/v1/submit/reject_activity

{
  "fingerprint": ""
}

List app proofs (audit)

Cryptographic proofs that an operation executed within Turnkey's secure enclave:

POST /public/v1/query/list_app_proofs

{
  "organizationId": "",
  "activityId": ""
}

Only available if the activity was created with generateAppProofs: true.

Three approver patterns

Pattern 1: Root user via Claude Code

The simplest approach. A human with root credentials uses Claude Code to review and approve pending activities.

Setup: None — root users can always approve.

Workflow:

1. List activities with CONSENSUS_NEEDED status
2. Review each activity's details (what action, who submitted, what parameters)
3. Approve or reject

Best for: Small teams, low volume, maximum human oversight.

Pattern 2: Non-root admin via Claude Code

A non-root user tagged 'admin' reviews and approves activities. Requires a policy whose consensus expression includes this user.

Setup:

1. Create the admin user with an 'admin' tag (see managing-users)
2. Create a policy whose consensus includes both the submitting agent AND the admin (see the submitter-in-consensus rule below)
3. The admin uses Claude Code with their own (non-root) credentials

Example policy requiring admin approval for large transfers:

{
  "policyName": "large-transfer-requires-admin",
  "effect": "EFFECT_ALLOW",
  "consensus": "approvers.any(user, user.tags.contains('')) && approvers.filter(user, user.tags.contains('')).count() >= 1",
  "condition": "activity.action == 'SIGN' && wallet.id == '' && eth.tx.value > 500000000000000000"
}

When an agent signs a transfer above 0.5 ETH, its auto-vote satisfies the agent-tag clause and the activity enters CONSENSUS_NEEDED. The admin then reviews and approves via Claude Code to satisfy the admin-tag clause.

` and are userTagId values returned by createusertag — the policy DSL compares against tag IDs, not the human-readable tagName. See the "Tag IDs vs. tag names" callout in managing-users`.

> Why both clauses? An ` count() >= 1 consensus alone would implicit-deny at submit time because the agent's vote contributes 0 to the admin count — the ALLOW never fires and the activity never reaches CONSENSUS_NEEDED. See [managing-policies → The submitter-in-consensus rule](../managing-policies/SKILL.md#the-submitter-in-consensus-rule). Requires the submitting agent to carry the agent tag (see managing-users`).

Best for: Teams that want human review without using root credentials.

Pattern 3: Automated agent approver

A dedicated agent that programmatically approves or rejects other agents' activities. This enables automated multi-party approval without human intervention for low-risk operations.

Setup:

1. Create an approver user with an 'approver' tag (see managing-users)
2. Create a policy on the worker agent that includes both the submitting worker and the approver in consensus, e.g. approvers.any(user, user.tags.contains('')) && approvers.filter(user, user.tags.contains('')).count() >= 1 (placeholders are userTagId values, not tag names — see [managing-policies → The submitter-in-consensus rule](../managing-policies/SKILL.md#the-submitter-in-consensus-rule))
3. The approver agent needs a narrow ALLOW policy — it should only be able to approve/reject activities, not sign transactions or create resources

Approver agent ALLOW policy:

{
  "policyName": "approver-can-approve",
  "effect": "EFFECT_ALLOW",
  "consensus": "approvers.any(user, user.tags.contains(''))",
  "condition": "activity.type in ['ACTIVITY_TYPE_APPROVE_ACTIVITY', 'ACTIVITY_TYPE_REJECT_ACTIVITY']"
}

Approver agent workflow:

1. Poll list_activities for CONSENSUS_NEEDED activities
2. For each pending activity, evaluate against approval criteria (amount thresholds, destination allowlists, activity type)
3. If criteria pass → approve_activity
4. If criteria fail → reject_activity or escalate to human

Security considerations for agent approvers:

• The approver agent must NOT be able to sign transactions itself — its only permission is approving/rejecting
• Define clear, programmatic approval criteria upfront (don't let the agent make judgment calls about unfamiliar activity types)
• High-value operations should still require human approval even with an agent approver — use a consensus threshold of 2 where one must be human
• Monitor the approver agent's own activity for anomalies

Best for: High-volume operations where human review of every transaction is impractical, but automated risk checks add value.

Combining patterns

You can require both human and agent approval. Include a clause for the submitting agent so the activity enters CONSENSUS_NEEDED instead of being denied at submit time:

{
  "consensus": "approvers.any(user, user.tags.contains('')) && approvers.filter(user, user.tags.contains('')).count() >= 1 && approvers.filter(user, user.tags.contains('')).count() >= 1"
}

This requires the submitting agent (first clause), one admin, AND one approver agent — the approver provides fast automated checks, the human provides judgment. Without the first clause the ALLOW would never fire on submit (see [managing-policies → The submitter-in-consensus rule](../managing-policies/SKILL.md#the-submitter-in-consensus-rule)).

For detailed setup guides and full request/response examples, see [references/approver-patterns.md](references/approver-patterns.md) and [references/activity-examples.md](references/activity-examples.md).

Troubleshooting

approve_activity fails with "no activity found with fingerprint" You're using the activityId instead of the fingerprint. Get the activity details with get_activity and use the fingerprint field.

Activity stuck in CONSENSUS_NEEDED Not enough authorized users have approved. Check the policy's consensus expression to see who can approve and how many are needed. Approvals expire after 24 hours.

Activity denied at submit time when CONSENSUS_NEEDED was expected The submitter isn't referenced by any clause in the ALLOW policy's consensus expression, so their auto-vote contributes nothing and the ALLOW never fires. Call get_policy_evaluations on the denied activity — you'll see the ALLOW with consensusMatched: false. Fix by adding a clause the submitter satisfies, e.g. approvers.any(user, user.tags.contains('')) && approvers.filter(user, user.tags.contains('')).count() >= 1 (substitute the actual userTagId values). See [managing-policies → The submitter-in-consensus rule](../managing-policies/SKILL.md#the-submitter-in-consensus-rule).

App proofs not available The activity must have been created with generateAppProofs: true. Proofs are not retroactively generated.

Agent approver can't approve Check that the approver agent's user tag matches the consensus expression in the policy. The approver also needs its own ALLOW policy for the APPROVE_ACTIVITY action.

Related Skills

• managing-policies — create policies with consensus expressions that trigger approval workflows
• managing-users — create admin and approver users with appropriate tags
• signing-transactions — activities generated by signing operations
• managing-agent — debug denied transactions with get_policy_evaluations

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: tkhq
• Source: tkhq/turnkey-agent-skills
• License: Apache-2.0

Install and usage instructions live in the source repository linked above.