Niches · Human approval & handoff

Hold a risky tool call until a person approves it

One company's docs show a full fit.

1companies, full fit
9companies, fit with a limit
16companies carry the tag
21share of the category's work
8first-hand accounts

What counts as done

The action does not run before an approval; the approver sees the exact payload; an approval runs that payload once, a rejection returns a typed refusal with the reason, and both reach the agent.

Measured, for our own tests: the decision reaches the agent within 10 seconds. No docs we read state these figures, so they don't count against a listing here.

How agents ask for it: ask me before you send that email, get approval before deleting anything, require sign-off for payments over 500

Does the job 1

The docs show it meets the test above, and nothing we read says otherwise.

ListingWhy
withHuman BB
Metoro Inc
Reviewers see the exact tool arguments; the decision is immutable, approved calls are single-use and must start before expires_at, and denials carry a note. GET with wait=30s returns the outcome, but no decision-to-agent latency figure is stated. source

Does the job, with a limit 9

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.

ListingWhy
gotoHuman D
gotoHuman UG (haftungsbeschränkt)
Reviewer sees the request data in a typed form; review.completed returns approved or rejected with the reviewed data. The hold is the builder's code (the MCP tool returns at once), there is no reason field unless added, and no delivery latency is stated. source
Hatchet B
Hatchet Technologies, Inc.
A durable task can pause on an event wait so the action placed after it does not run first. Showing the approver the payload, the refusal type and delivery are the builder's code, and no resume latency is stated. source
Inngest BB
Inngest Inc
step.waitForEvent or waitForSignal holds the run so the side effect after it waits; the guide's pattern returns rejected with a reason. Payload display and delivery are the builder's code, no resume latency is stated, and issue 2423 reported a 3.5-minute resume (closed). source
Orkes Conductor Human tasks C
Orkes, Inc.
A Human task pauses the workflow until a form is submitted in the Conductor UI or Human Tasks API; read-only fields show the payload and the form response becomes the output. Human tasks are off until Orkes support enables them, and no resume latency is stated. source
Permit MCP Gateway C
Permit Inc.
The gateway pauses the tool call, shows admins the exact arguments, runs that call on approval and returns an error with the admin's reason on rejection, outside the model. Enterprise plan only, and no decision-to-agent latency is stated. source
Pushary C
RalphNex OÜ
Host permission hooks (Claude Code, Codex, Cursor, Gemini CLI) hold the tool call and send the full command as actionBody; deny returns to the agent. Plain MCP clients get cooperative questions only, the map's named blocker, and no decision latency is stated. source
Restate BB
Restate GmbH
A handler awaits an awakeable before the action; resolve continues it and reject sends a terminal error with the reason. Sending the payload to the approver is the builder's code and no resume latency is stated. source
Temporal BB
Temporal Technologies Inc.
A workflow blocks on a Signal with a typed decision, approver and comments before the next step. The approval interface is the builder's, Signals are fire-and-forget unless an Update is used, and no decision-to-resume latency is stated. source
Trigger.dev B
API Hero Ltd
wait.forToken pauses the run until the token is completed with typed output such as approved or rejected; an idempotency key stops a retry sending a second request. The reviewer UI and payload display are the builder's, and issue 2295 saw an 11-minute resume (closed 2025). source

Doesn't do this job 1

Listings in the category that do something else, with the reason.

ListingWhy
Big Arrow BB
Franz Enzenhofer (individual, Vienna)
A card --button returns the person's choice to the agent, but nothing holds the action: the agent runs it itself afterwards, there is no payload binding and no typed refusal with a reason. source

What stops agents finishing

  • approval step only on Enterprise or contact-sales plans
  • plain MCP clients get cooperative questions only, so the agent can skip the gate
  • side effect placed before the wait in the workflow

Around it

Before: guardrails.classify-risk. After: email.send, payment-platforms.charge, code.merge-pr.

Where this comes from

Weight 21 of 100 in Human approval & handoff: 8 of 26 accounts (Permission Slip, Call a Human MCP, n8n thread, nono, withHuman, HumanLayer); the most described job

The same as data: /api/v1/use-cases/human-in-the-loop.approve-action.json, or the MCP tool get_use_case with id human-in-the-loop.approve-action. Other jobs in this category: /api/v1/niches?category=human-in-the-loop.

Draft, read 2026-11-15 from public docs and first-hand accounts, not yet reviewed by a person. A mark says what the docs show, not what a run measured.

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.