Using the Dashboard

Overview

The Agent OS dashboard is the web UI for managing agents, tasks, skills, graphs, and settings. It is typically served from the same origin as the API (e.g. http://127.0.0.1:3010) or from a separate static build that points to the API. Use it to create agents, run tasks, install and configure skills, edit graphs, and set API keys and channel credentials.

Main areas

Agents

• List agents — See all agents (id, name, model, skills, schedule). Create, edit, or delete agents.
• Run / chat — Select an agent and run a one-off task or chat in a conversation. Streaming responses and tool calls are shown when supported.
• Schedule — Set or edit cron schedule and optional schedule_input per agent for scheduled runs.
• Workspace / identity — View or edit agent workspace files (e.g. IDENTITY.md, USER.md, SYSTEM.md) if the UI exposes them.
• Manage memory — Link to the Memory page with agent pre-selected for that agent’s memories.

Tasks

• Task list — View recent tasks (status, agent, graph, timestamps). Filter by agent_id, graph_id, status.
• Task detail — Open a task to see logs and result.

Skills

• List installed skills — See skills loaded from ~/.agent-os/skills/.
• Install system skills — Copy default skills from the seed dir into ~/.agent-os/skills/.
• Install from store — If a registry URL is configured, browse and install skills by slug.
• Upload — Upload a skill (zip or Markdown) if the UI supports it.
• Config — Per-skill config (API keys, options). Schema comes from the skill’s config.schema.json; values stored in ~/.agent-os/configs/<skill_id>.json.
• Setup — View SETUP.md for a skill (how to get credentials, etc.).
• Uninstall — Remove a skill from ~/.agent-os/skills/.

Graphs

• List graphs — See saved workflows (nodes = agents, edges = flow).
• Edit graph — Add/remove nodes and edges, set schedule and schedule_input.
• Run graph — Execute a graph (by id) with optional input; view streaming or final result.

Logs

• Recent logs — View agent/task logs (from the server’s log store or events). May be combined with WebSocket for live updates.

Memory

• List / search memories — By agent, user, or query. Optional semantic search if enabled.
• Add memory — Create a memory entry (agent_id, user_id, content).
• Delete memory — Remove a memory by id (e.g. from the list or via API).

Settings

• AI / API keys — Set OpenAI, OpenRouter, Anthropic, Google API keys (or other providers). Stored in ~/.agent-os/config.json.
• Channels — Configure Telegram, Slack, Discord, Signal, Viber: bot token, signing secret or public key, default agent. Each channel may have a “Configure” flow and status.
• Store / registry — If supported, set the skill store registry URL for “Install from store”.
• Other — Any other server or UI settings exposed by the build.

Tips

• Agents not showing? Run sulala onboard or ensure the database exists and agents are seeded (or create an agent manually).
• Skills not loading? Ensure skills are under ~/.agent-os/skills/ and have skill.yaml or SKILL.md. Restart the server or use Install system skills.
• Channels not receiving messages? Check token and default agent in Settings → Channels. For webhooks, ensure the server is reachable over HTTPS (e.g. ngrok) and the webhook URL is set on the platform (Telegram, Viber: use set-webhook endpoint).
• Dashboard not loading? Ensure the server is running (bun run dev or sulala start) and you’re opening the correct URL (default http://127.0.0.1:3010).