Human-in-the-loop API for builders
Add human approval to your AI product.
Reach your customer in the native Pushary app for a confirmation, choice or written answer. Pushary handles enrollment, delivery and the recorded response. Your application protects the action, validates the answer and continues the correct workflow.
Your product pays. Enrolled customers answer in the native app without a separate account or paid plan.
How it works
From customer question to validated answer
Your backend creates the review, the enrolled customer answers on their phone, and your application validates the outcome before continuing.
- 01
Your agent creates a decision
Your backend selects the authorized reviewer, completes customer enrollment, and creates a confirm, select or input decision. Keep protected actions behind an enforced application or framework gate; an optional ask tool does not enforce approval.
- 02
Your end-user approves, denies, or answers
The enrolled customer answers in the native Pushary app. Confirm requests can offer notification actions; choices and written answers open the app. Slack and legacy web compatibility remain available where configured. Your product pays; the reviewer needs no separate Pushary account or paid plan.
- 03
You resume on the answer
Your backend reads the recorded outcome after a signed callback or polling. Validate the intended reviewer, action and saved run before continuing. Your application owns persistence, replay handling and recovery if continuation fails or its result is uncertain.
Decisions come in three shapes: a yes or no confirm, a multiple-choice select, or a free-text input. Every one is stored with its outcome and kept in the audit log for a year.
What you get
The approval plumbing, done once
Customer enrollment, phone delivery and recorded answers are built in. Your backend verifies callbacks and owns protected execution and recovery.
Durable decisions API
Each decision has a stored ID, status and outcome that your backend can read after the original request ends. To continue a stopped workflow, your application must also save its own run state and decision mapping. A recorded answer is not proof that the action completed.
Signed webhooks
A signed callback tells your backend about a decision outcome. Verify it using the documented signing scheme, then validate the stored decision against the intended reviewer and action. Signature verification alone does not authorize execution or prevent duplicate processing.
TypeScript SDK, REST, and MCP
Call it however your stack works. The typed SDK for Node and edge runtimes, a plain REST endpoint for any language, or the MCP tool so an agent framework can create a decision as a native tool call.
365-day audit log
Who asked, what was asked, who answered, and when, kept for a full year and exportable. When a customer or an auditor asks why your product did something, you have the record.
Native customer answers
Enroll the customer in the native Pushary app and ask for confirmation, a choice or text. Confirm can offer notification actions; select and input open the app. Existing Slack delivery and legacy web compatibility remain available where configured.
Scoped to your product
Resolve the authorized reviewer from your application, then bind the decision to that enrolled customer. Present your requesting-product identity within Pushary. This does not require a private-label mobile app, and the person requesting an action is not automatically its approver.
Who builds on it
Teams putting an agent in front of real users, where a wrong autonomous action has a cost and a human sign-off is expected.
AI product teams
You ship an agent that needs customer input before proceeding. Use Pushary for enrollment, delivery and recorded answers while your application keeps its permissions, execution and run storage.
Automation and workflow platforms
Your users build flows that need customer answers. Connect a recorded response to the framework's pause and resume mechanism, with persistent state and explicit outcome handling in your application.
Vertical SaaS with agents inside
Your product needs a named reviewer and a record of their answer. Start with a compact decision the customer can understand on their phone; your team remains responsible for domain-specific permissions and requirements.
Agent framework builders
You maintain a framework or template others build on. Offer a first-class human-in-the-loop primitive your users can call over MCP or REST instead of asking them to roll their own.
Pricing
One flat plan for the embed
Embed the approval loop in your own product for your own end-users. No per-seat math on your customers.
- Up to 25,000 end-users
- 100,000 decisions per month
- Durable decisions API, signed webhooks
- TypeScript SDK, REST, and MCP access
- 365-day exportable audit log
- Native customer answers
FAQ
Questions, answered
How do I add human approval to my AI agent?
What is a human-in-the-loop API?
Can my serverless agent wait for a human?
How do I verify the webhook?
How is this different from notifications?
Does my end-user need a Pushary account?
What does the AI agent approval SDK support?
How much does the embed use case cost?
Give your product a way to ask a human.
Enroll your customer, send a compact question to their phone, and retrieve the recorded answer. Your application validates it and continues the intended workflow.
Start with PartnerWant the code first? Read the integration guide.