Overview
What your agents are doing right now.
Agent Activity
Last 7 daysTeam Map
Recent Agent Runs
View all runsNeeds attention
Subscription limits
—
| Task | Kind | Assignee | Status | Progress | Repo | Updated |
|---|
Connect your Claude & Codex subscriptions once. Agents can't run until at least one engine is verified.
Run tasks automatically — on a cron, on a repeating heartbeat, or on demand via a webhook.
A goal + a project manager that autonomously breaks it into tasks and delegates to your teams & agents.
Project groups super-categories that share variables (e.g. one n8n instance) — inherited by every project in the group
Define agents and assign their MCPs, skills, plugins, and secrets.
An orchestrator plus members. Team-wide MCPs/skills/plugins merge into every member at runtime.
Team Map
All teams
Reusable building blocks — available to tick on any agent or team.
MCP servers
Skills SKILL.md units — self-written or from the host; may bundle scripts & files
Plugins installable bundles — skills, commands, agents & MCPs from marketplaces & GitHub
Connection, appearance, and health.
API token
Stored in this browser. Required when the daemon has AGENCY_API_TOKEN set.
Appearance
Follows your system by default. Cycle with the toggle in the sidebar, or pick here.
Health
Usage estimated tokens & runs by day · agent · engine
Infisical Machine Identity fetches per-customer secrets live from your vault
Paste a least-privilege, read-only Machine Identity. The client secret is stored 0600 on the server and never shown again.
Customer → Infisical project
Map a customer slug to its Infisical project so any repo (or a general task) can pull its secrets from just the Customer field — no kunden/<kunde>/.infisical.json needed. A repo that does carry that link still wins.
Connect an external agent (MCP) give Claude Code / Hermes your full orchestration rights
Agent-Hub exposes an MCP server. Connect a client and it can list/create & delegate tasks to your agents and teams, approve/steer running tasks, and manage agents/teams — the same powers you have here.
Secrets
Values are written to secrets.env on the server and injected into agents/MCPs at runtime. Values are never shown back here.