Build Agents

Spawnbase gives you two ways to build agents: describe what you want in the copilot chat, or drag-and-drop nodes on the visual canvas. Most people start with the copilot and refine on the canvas.

The Copilot

Open the copilot panel (press A or click the chat icon in the footer) and describe what you want to automate:

Every morning at 9am, pull unresolved Sentry issues from the last 24 hours,
group them by root cause, rank by severity, and post a summary to #engineering in Slack

The copilot understands your intent, plans the workflow structure, and builds the nodes on the canvas. You review the plan, approve it, and the agent is ready to configure and deploy.

The copilot can also modify existing agents — ask it to add a step, change a schedule, or swap an integration.

The Canvas

The canvas is a visual workspace where your agent’s nodes and connections live. Each node is a step in the workflow, and edges show how data flows between them.

Working with the canvas:

Add nodes — press N to open the node picker, or right-click the canvas
Connect nodes — drag from a node’s output handle (right) to another node’s input handle (left)
Configure nodes — click any node to open its configuration panel on the left
Pan and zoom — scroll to zoom, drag the canvas to pan (or switch to select mode with V)
Annotate — add sticky notes to document your workflow logic

Node Types

Every agent is built from four types of nodes:

Triggers — When the agent runs

Every agent starts with a trigger. It determines when and how the agent executes.

Trigger	What it does
Schedule	Runs on a cron schedule — hourly, daily, weekly, or custom
Manual	Runs on demand via button click or API call

AI Agent — Reasoning and decisions

The AI Agent node is what makes Spawnbase agents intelligent. It calls an LLM (Claude, GPT-4o, Gemini, Grok, and others) to reason about data, make decisions, and generate output.

Without tools, it makes a single LLM call — good for summarizing, classifying, or generating text. With tools enabled (web search, code sandbox, MCP servers), it becomes an autonomous agent that reasons, acts, and iterates until the task is done.

App Actions — Connecting to your tools

App action nodes call external services — send a Slack message, create a Linear issue, query PostHog analytics, or any of 25+ supported apps.

You pick an app, connect your account (OAuth or API key), choose a tool, and map inputs from upstream nodes.

Annotations

Sticky notes for documenting your workflow. Non-executable — they’re just for you and your team.

Data Flow

Data passes between nodes through template variables. Each node’s output is available to downstream nodes:

{{schedule_trigger.output}}
{{sentry_issues.result}}
{{ai_summary.text}}

Click any node to see its output schema and available fields. The configuration panel shows which variables you can reference.

Testing and Execution

Before deploying, test your agent to make sure it works:

Click Run in the header
Watch execution in real-time in the bottom panel — each node lights up as it runs
Click any node during or after a run to inspect its input/output
Fix issues and re-run

The execution panel shows timing, status, and full I/O for every step. If a node fails, you see the exact error and can fix it without guessing.

Keyboard Shortcuts

Shortcut	Action
`A`	Open copilot chat
`N`	Open node picker
`V`	Switch to select mode
`H`	Switch to pan mode
`Backspace`	Delete selected node
`Cmd + Z`	Undo
`?`	Show all shortcuts