Loading pattern preview
BlocksHuman-in-the-Loop
HIL Tool Approval Basic
HIL Tool Approval Basic
Chat assistant with human approval workflows for safe tool execution.
Use this when you want a basic starting point for gating risky agent tool calls behind a human approve or deny step, with Upstash in the stack.
Loading pattern preview
Problems solved
Technical challenges this implementation handles out of the box.
- Intercept tool calls that require human confirmation before execution
- Resume the agent after approve or deny decisions
- Limit agent loops with stepCountIs during approval flows
- Surface pending tool approvals in a streaming chat UI
Use cases
Products and workflows this pattern is designed to support.
- Content moderation queues with human sign-off
- E-commerce actions that need cart confirmation
- Admin tools that gate destructive operations
- Customer-support agents requiring human approval
Compare with
Similar patterns, and when each one is the better fit.
HIL Needs ApprovalDemonstrate tool approval with async generators. Includes real-time feedback, approval workflows, and interactive management.
Tool ApprovalUse this when a destructive action, such as deleting a file, needs an in-chat confirm or deny step built only on the AI SDK's own approval flow.
HIL Inquire Text InputUse this when your agent needs to pause mid-loop, ask the user an open-ended question, and resume once they type a free-text answer.
HIL Inquire Multiple ChoiceUse this when an agent loop must stop on an inquire tool call to collect a multiple-choice answer before continuing with that context.
Setup
Requirements, wiring steps, and what this pattern adds to your project.
Getting started
Pick how you want to pull this pattern in. Then wire env vars and routes the same way.
- 1
Download → Next.js app
Use Download → Next.js app in the preview toolbar. You get a zip scaffold (Pro downloads rewrite preview `/view/.../api/*` paths to `/api/*` automatically).
- 2
Extract the zip and install dependencies
Unzip the archive, then run `pnpm install` (or npm / yarn) in the project folder.
- 3
Fill .env.local
The download includes .env.local placeholders when needed. Add real values for: OPENAI_API_KEY.
- 4
Customize the agent and tool files
Adapt prompts, tools, and stop conditions in the generated project for your product.
- 5
Run the app and open the pattern route
Run `pnpm dev`, open the setup/checklist page if present, then navigate to the generated pattern route and verify responses.
Environment variables
OPENAI_API_KEYGet key
Capabilities
AI SDK APIs
streamTextgatewaystepCountIsconvertToModelMessagestoolProviders
OpenAI
External services
Upstash
Files added
6 files
app/page.tsxapp/layout.tsxapp/api/human-in-loop/route.tslib/tools.tslib/utils.tsREADME.md
Dependencies
8 total
npm packages
6AISDKlucide-reactNext.jsReactsonnerZod
Registry components
buttonempty
Critical files
3 files
human-in-loop/route.tsAPI route implementing the human-in-the-loop pattern. Uses processToolCalls() to intercept tool calls that require human confirmation before execution. Passes executor functions for tools without execute. Uses stepCountIs(5) to limit agent loops.
lib/tools.tsDefines 7 AI tools demonstrating the human-in-the-loop pattern. 5 tools lack execute functions (requiring human approval): getWeatherInformation, moderateContent, addToCart, approveUser, processPayment. 2 tools have execute functions (auto-execute): webSearch, getLocalTime.
lib/utils.tsCore utility implementing the processToolCalls pattern for human-in-the-loop flows. Processes tool parts from the last message, checks for APPROVAL.YES/NO strings, executes approved tools via provided executor functions, and streams results back via UIMessageStreamWriter.
Requirements
- Complexity
- Intermediate
- Setup time
- ~15 minutes
- Node
- 24+
- Memory
- 512MB
Upstash Redis
