Your first workspace
Build a flow from scratch in three minutes and watch the agent work.
Your first workspace
We'll build a minimal flow: a code editor, a terminal, and a Claude agent with access to both. At the end you'll ask the agent to read a file in your project and run something in the terminal — and it'll do it, with you watching every step.
1. Create the workspace
- Home screen → Create.
- Pick a name and a folder. The
.riggrfile holds the whole canvas (nodes, connections, viewport). - The empty canvas opens. It's an infinite space you navigate with right-click + drag, W A S D, or arrow keys.
2. Add three nodes
Quick search is the way — forget menus.
- Ctrl+K (or ⌘K on Mac) → type
code→ add Code. - Ctrl+K →
terminal→ add Terminal. - Ctrl+K →
agent→ add Agent Terminal.
Three blocks on the canvas. Drag each into a comfortable spot.
3. Wire the nodes into the agent
Every node has handles on its edges — connection points.
- Click a handle on the Agent Terminal and drag it to a handle on Code.
- Repeat to wire the Agent Terminal to the Terminal.
Two lines, two potential accesses. We still need to say what the agent is allowed to do with each one.
4. Set the permissions
In the middle of each line sits a chip — a tiny button that represents the connector for that link.
- Click the chip on Agent → Code. Pick code-read-write (read + write) or code-full-access (also opens files).
- Click the chip on Agent → Terminal. Pick execute to allow command execution.
Done. The agent has scoped access.
5. Start the agent
Click inside the Agent Terminal and press Start (or ▶). Riggr:
- Spins up a local MCP server for the node.
- Spawns the
claude-codeconfigured in Settings. - Injects into the system prompt the list of connected nodes (Code and Terminal in our case).
A few seconds later you see Claude Code's prompt inside the embedded terminal.
6. Talk to it
Send something concrete:
Read my project's `package.json` and explain the scripts.
Then run `npm test` and show me the result.
The agent reads through the Code node, runs through the Terminal, and answers — all visible on the canvas, in real time. Its tool calls show up as chat lines (> Tool: riggr_read_code_node_1).
7. Save
The workspace autosaves 500ms after the last change (configurable in Settings → Workspace). Ctrl+S forces a save now. To reopen later, double-click the .riggr or open it via Riggr.
Where to go next
- Node catalog — every one of the 16 blocks, with config props and examples.
- Connections and permissions — deep dive on the chip system.
- Typed ports — how data flows between nodes without going through the agent.
- Examples — five ready-made workspaces you can adapt.