Security

Security & Trust Overview

What the bot can and cannot do inside your community, what data it touches, and where that data goes.

Last updated: July 2026
The short version: HexTriage is a read-and-reply support bot. It holds no funds, executes no transactions, and requests no wallet permissions. It runs with the minimum platform permissions needed to answer questions in designated channels, and cannot initiate contact with your users by design.
Least Privilege

Platform permissions

Access is scoped to exactly what's needed to answer questions, nothing more.

Discord

  • ✓ Read Messages, Send Messages, Read Message History, Embed Links, in designated support channels only
  • ✕ No Administrator
  • ✕ No Manage Server
  • ✕ No Manage Roles
  • ✕ No Manage Webhooks

Telegram

  • ✓ Standard Bot API
  • ✓ Reads and replies only in groups it's added to
  • ✕ No admin rights

Access is scoped per channel and revocable instantly.

By Design

Anti-phishing posture

  • ✓ Never initiates DMs, since scam bots DM first and HexTriage structurally cannot
  • ✓ Never asks for private keys, seed phrases, or wallet connections, and warns users that legitimate support never will
  • ✓ No signing capability of any kind
  • ✓ Wealth-blind, meaning users are never profiled or prioritized by wallet balance
Data Handling

Data processed and stored

Data Purpose Retention
Conversations (user question + bot answer) Quality assurance: catching and correcting bot mistakes 30 days, auto-deleted
Flagged or escalated conversations Extended quality review 90 days, auto-deleted
Transaction hashes & contract addresses Failed-transaction diagnostics 30 days
Platform user IDs Message delivery while a ticket or session is open Deleted when it closes
Aggregate statistics (counts, categories, deflection rate) Monthly reporting to your team Contract duration, contains no personal data

User identity in conversation logs is stored only as an irreversible cryptographic hash. Raw usernames are never written to conversation records. No emails, KYC data, or off-chain personal identifiers are ever collected.

Transparency

Transcript transparency

Clients can review their own community's conversations in their dashboard at any time, strictly isolated per client, with one-click flagging of any answer for review. Your bot's work is fully auditable, no black box.

Where Data Goes

Subprocessors

AI: Anthropic (Claude API)

Receives message text for response generation. Per Anthropic's commercial terms, API inputs and outputs are not used to train models.

Blockchain data

Industry-standard on-chain data and contract-intelligence providers, which receive only public identifiers (transaction hashes, contract addresses), data already visible on any block explorer. No message content is ever shared with them.

Security screening

A third-party token-security database, receiving contract addresses only.

A complete, named subprocessor list is available to clients on request as part of our Data Processing Agreement.

Infrastructure

How it's hosted

  • ✓ Hosted in the EU (Amsterdam, Netherlands)
  • ✓ Credentials stored in managed secret configuration, never in code or repositories
  • ✓ All services served over SSL
  • ✓ Rate limiting protects communities during spam floods and incident surges
  • ✓ A 115-test automated suite gates every deployment
Compliance

Compliance and data rights

EU-operated with GDPR-aware practices: data minimization, hashed identifiers, defined retention, deletion on request. A Data Processing Agreement is available on request. Client knowledge bases are fully isolated, so one protocol's documentation never informs another's bot.

Incident Response

Your kill switch

Security contact

Reach us at security@hextriage.io, acknowledged within 24 hours.

Ultimate control is yours: revoking the bot's channel permissions ends all access instantly, no action from us required.