- BLOCK
- ALLOW
- REVIEW
- LOG
Short answer
How can I apply manual risk decisions with my AI?
Your agent uses Sift tools to view pending decisions and execute them on specific accounts. You can instruct your agent to block or allow users based on the risk data it sees. This allows you to handle fraud response through natural language.
Where every decision lands.
Risk management outcomes
The agent processes your commands into specific Sift actions.
BLOCK
User is restricted
The agent applies a block to a suspicious account. This stops further fraudulent activity immediately.
ALLOW
User is cleared
The agent lifts restrictions on an account. This restores access for legitimate customers caught in a false positive.
REVIEW
Human intervention
If the risk is unclear, the agent flags the case for you. You decide the final outcome after checking the details.
LOG
Audit trail updated
Every action the agent takes is recorded in Sift. You get a clean history of why a user was blocked or allowed.
The workflow
What your AI does when a risk alert arrives.
The agent bridges the gap between seeing a risk and fixing it.
Fetch risk data
The agent pulls the current list of pending decisions from your Sift account to see what needs attention.
list_sift_decisionsAnalyze the user
Your agent looks at the specific risk signals and user history to suggest the best course of action.
list_sift_decisionsExecute the decision
Once you confirm, the agent tells Sift to apply the specific decision to the user's profile.
apply_user_decisionConfirm completion
The agent verifies the decision was successful and updates your chat so you can move to the next task.
apply_user_decision
Try it
Copy these to start.
Use these prompts to trigger actions in Sift.
Starting points
Swap out the bracketed info with your actual user IDs or decision types.
Sift Connector
You're all set. Choose your MCP client and follow the setup instructions.
https://edge.vinkius.com/vk_preview_X6jrj9G3NJ6qSu4scGacFQiORtOckhn98ppSzoMs/mcpClaude Desktop
Follow the steps below to connect in seconds.
- 1In Claude Desktop, open Settings → Connectors.
- 2Click “Add custom connector” and paste the connector link above as the remote MCP server URL.
- 3Click Add and start a new chat — Sift capabilities are ready to use.
{
"mcpServers": {
"sift-chargeback-mcp": {
"url": "https://edge.vinkius.com/vk_preview_X6jrj9G3NJ6qSu4scGacFQiORtOckhn98ppSzoMs/mcp"
}
}
}List all the pending Sift decisions that need my review.
Apply a block decision to user ID 98765.
Check the current risk status for account user_abc_123.
Allow the user with ID 55443 to proceed with their transaction.
Claude
ChatGPT
Cursor
VS Code
Windsurf
Claude Code
JetBrains
Cline
Start here
Connect Sift once, then ask.
Just link your Sift account to Vinkius. Your credentials stay encrypted and the agent handles the rest.
Connect Sift to your AIFAQ
How this task behaves
- 01
Can the AI block users automatically without me?
No. The agent only applies decisions when you explicitly instruct it to do so in chat.
- 02
Can the AI delete user data from Sift?
No. The tools only allow for applying specific risk decisions like blocking or allowing users.
- 03
Does the AI see my Sift login password?
No. Vinkius manages the connection securely so your agent only uses the authorized tools.
- 04
Can I undo a decision the AI made?
Yes. You can tell the agent to apply a different decision to the same user to reverse an action.
- 05
What happens if the Sift API is down?
The agent will tell you it cannot reach Sift and won't be able to pull or apply any decisions.
More questions about Sift? See everything the Sift Connector can do
Connect Sift to Claude, Cursor, ChatGPT & more
