Niches · Human approval & handoff
Ask a person a question mid-task and carry on with the answer
Two or three companies' docs show a full fit.
What counts as done
The question reaches the right person, the answer comes back typed (an option or text) with who answered, and the agent uses it in its next step; with no answer it does not proceed on its own guess unless told to.
How agents ask for it: which customer did you mean
, ask finance whether to refund
, check with me if anything's unclear
Does the job 2
The docs show it meets the test above, and nothing we read says otherwise.
| Listing | Why |
|---|---|
| gotoHuman D gotoHuman UG (haftungsbeschränkt) | Review types take dropdown, checkbox, button and text input fields; the webhook returns the filled data with respondingUser and respondedAt. assignTo routes to named reviewers (Team plan and above). source |
| Orkes Conductor Human tasks C Orkes, Inc. | User forms collect typed fields, validated on submit, and the form response is the task output; assignment targets a user or group, and the trigger context carries assignee and claimant. source |
Does the job, with a limit 7
It does the job, but something in the docs or a user's account will get in an agent's way. The limit is named in each row.
| Listing | Why |
|---|---|
| Big Arrow BB Franz Enzenhofer (individual, Vienna) | --ask (typed reply) and --choice (2 to 6 options) return JSON answers by row id, and exit 1 when the card is closed unanswered. The answerer is whoever sits at that Mac; no identity is returned. source |
| Hatchet B Hatchet Technologies, Inc. | The event payload can carry any typed answer and who gave it, but Hatchet neither asks the person nor records the answerer; both are the builder's code. source |
| Inngest BB Inngest Inc | The resuming event or signal can carry any typed answer and the answerer, but Inngest neither asks the person nor records who answered; both are builder code. source |
| Pushary C RalphNex OÜ | ask_user takes confirm, select (2 to 6 options) or input and returns value and note with a status; expired is documented as not approval. Teams record the answerer by name, but the tool result is not shown to carry who answered. source |
| Restate BB Restate GmbH | The resolution carries any JSON answer, so options or text work, but Restate neither asks a person nor records who resolved it; both are builder code. source |
| Temporal BB Temporal Technologies Inc. | A Signal or Update can carry any typed answer and the approver's identity, but Temporal neither asks the person nor verifies who answered; the page tells builders to validate approver permissions themselves. source |
| Trigger.dev B API Hero Ltd | Token output is any typed JSON, so options or text work, but the person is reached by builder code and nothing records who completed the token. source |
Doesn't do this job 2
Listings in the category that do something else, with the reason.
| Listing | Why |
|---|---|
| Permit MCP Gateway C Permit Inc. | Approve or reject only, with a rejection reason; there is no question, option or free-text answer type. source |
| withHuman BB Metoro Inc | Approve or deny only, with an optional note; issue 144 asking for a question tool is still open. source |
What stops agents finishing
- approval-only product with no question type
- question routed to the operator when someone else holds the answer
- short wait cap returns before the person can answer
Around it
Before: agent-harnesses.run-agent. After: human-in-the-loop.approve-action.
Where this comes from
Weight 11 of 100 in Human approval & handoff: 4 of 26 accounts (Call a Human MCP, TTal mid-job blockers, Claude Code AskUserQuestion, withHuman ask_human request)
The same as data: /api/v1/use-cases/human-in-the-loop.ask-question.json, or the MCP tool get_use_case with id human-in-the-loop.ask-question. Other jobs in this category: /api/v1/niches?category=human-in-the-loop.
For companies
Do agents find, use and choose your tools?
An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.