> For the complete documentation index, see [llms.txt](https://legacy-docs.usesmileid.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://legacy-docs.usesmileid.com/further-reading/faqs/agent-mode.md).
# What is agent mode?
## Agent Mode
Agent mode is a feature that enables your agents on the field to easily capture selfies of users they are onboarding. This feature is designed to increase the pass rate of jobs.
#### How it works
If a device has a secondary camera, and you enable the setting to allow-agent mode in your configuration code; your agent will have the ability to switch the camera to the back camera to capture user selfies.
* Mobile devices tend to have a selfie and back cameras. Agent mode will automatically pick the most appropriate back Camera.
* On desktop/laptop computers, this will depend on whether a web-cam or [continuity camera](https://support.apple.com/en-gb/102546) is attached to your device.
> `allowAgentMode:true` on mobile and `allow-agent-mode`: `true` on web
#### Availability
1. Mobile SDKs (ios, android, flutter, react-native, expo): v10 & Above
2. Web Client SDKs (web-embed/hosted-web, javascript sdk): v1.4 >
#### Best Practices
Train your agents on how to capture the best images of users and how to use agent mode in your app. For best results, ensure the agent captures users where background is plain and the environment is well-lit. Too many people, Shadows or backlighting can affect image quality and verification accuracy.
**1. Prepare the Device**
Before beginning any capture, ensure that the device and environment are optimal for high-quality images.
Wipe the lens clean before you begin. A clean lens ensures there are no smudges or dirt that could affect image clarity.
Ensure sufficient battery and camera permissions. Running out of battery or lacking camera access can disrupt the process and compromise results.
**2. Positioning & Framing**
Proper positioning ensures the best chance for accurate capture and reduces rejections due to framing errors.
Hold the device at eye level, about arm’s length from the user. This helps avoid distortion and captures a true likeness.
Keep the user centered; avoid tilting or shaking. The subject should remain within the frame, with their face clearly visible and not cut off at any edge.
**3. Guiding the User**
Clear instructions help the user cooperate, resulting in better images and fewer retakes.
Explain each prompt clearly and read them aloud as they appear. This ensures users know what is expected at every step.
Ask the user to remove glasses (if they are able to) hats, or face coverings. Obstructions can interfere with face detection and verification accuracy.
**4. Monitoring & Feedback**
Observe and respond during the capture process to ensure successful completion.
Watch for the blue mesh to confirm detection. The mesh indicates the system has detected a face and is ready to proceed.
Observe the green progress stroke—if it stalls, re-center or remind the user to stay still. Stalled progress usually means movement or incorrect positioning.
**5. Review & Submission**
Final checks help prevent errors and avoid unnecessary re-submissions.
Compare the preview image with the user for clarity. Check that the image is sharp, well-lit, and correctly framed.
If it’s not sharp, select Retake and repeat the flow. Always prioritize quality as blurry or incomplete images reduce the chance of successful verification.
#### Agent Mode FAQs
**Q1. What is Agent Mode and how do I activate it?**
A: Agent Mode is the camera workflow for onboarding. On the mobile device, toggle Agent Mode on, then tap Start Capture to launch the live camera view.
**Q2. How do I know when the user’s face is detected?**
A: Point the camera to the user and as soon as the face is in view, a blue mesh overlay will appear over the face to confirm detection.
**Q3. What do the on-screen prompts and progress indicator mean?**
A: Prompts appear below the video frame (e.g., “Center your face,” “Hold still,” etc.). Each time the user completes a prompt successfully, a green stroke animates around the frame to show progress.
**Q4. When is the smile capture performed?**
A: After initial poses, you will see/read out a “Please smile” prompt. The system detects a smile once the mouth-mesh stretches a set distance (teeth visible).
**Q5. What happens after a successful smile capture?**
A: A preview screen displays the frozen image. Both agent and user review it—either Confirm to submit or Retake to restart the flow.
**Q6. What if the capture is blurry or unclear?**
A: Hold your phone still while performing the capture. Stop and clean the camera lens. Blurry images can lead to failed detection or poor quality submissions.
**Q7. Can more than one person be in the frame?**
A: Multi-person captures are unsupported. Ensure only the onboarding user is in view.
**Q8. What environment works best for the capture?**
A: A plain, uncluttered background with even lighting. Avoid hats, masks, or anything that covers the user's face.
#### Issues
Report any issues you encounter with this feature to us via the github repos. Search our [known issues](/further-reading/known-issues.md) file for current logged issues.
---
# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.
## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.
Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:
```
GET https://legacy-docs.usesmileid.com/further-reading/faqs/agent-mode.md?ask=&goal=
```
`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.
Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.