Clavis: AI Agent Command Center
Clavis is an AI agent command center (Next.js + TypeScript) that lets an agent act across 4 third-party services (Gmail, GitHub, Slack, Discord) via Auth0 Token Vault for secure, zero-exposure credentials, fronted by a centralized risk engine that governs what the agent is allowed to do.
One Command Center, Four Services
Clavis is the only AI command center that bridges Gmail, GitHub, Slack, and Discord without ever touching your passwords. Instead of jumping between apps, you issue natural-language instructions and the agent decides which of 12 tools to call across the four connected services.
Secure Credential Management
Every integration authenticates through Auth0 Token Vault using Connected Accounts, so the app obtains short-lived access tokens rather than storing raw credentials. This zero-exposure design means sensitive OAuth tokens are never held by the client or persisted in the app.
Risk Engine & Step-Up Auth
A centralized risk engine categorizes each action as low, medium, high, or critical. Medium-risk operations (primarily write actions like posting a Slack message or creating a GitHub issue) require step-up authentication before they run — a secure-by-default approach to letting an agent act on your behalf.
Auditing & Hard-Won Lessons
A persistent audit log captures timestamps, requested scopes, and outcomes for accountability. Building it surfaced real integration lessons documented in the repo: Token Vault needs Connected Accounts (not Account Linking), Slack's lack of refresh tokens forced a bot-token approach, in-memory logging was replaced with persistent storage, and natural-language inputs had to be validated against each API's requirements.
Key Features
Natural-language control of Gmail, GitHub, Slack, and Discord from one command center.
12 integrated tools: email search & calendar (Google); repos, issues & profile (GitHub); channels, history & posting (Slack); profile, servers & membership (Discord).
Risk engine classifies every action as low / medium / high / critical, and gates medium-risk writes behind step-up authentication.
Zero-exposure credentials via Auth0 Token Vault — the agent uses short-lived access tokens and never touches raw passwords.
Persistent audit log records timestamps, requested scopes, and action outcomes for full accountability.