Building an All-in-One Slipbox Agent

Slipbox is really designed to be read and acted on by an agent rather than browsed by a person directly. In this tutorial, you'll connect all 11 Slipbox nodes to an Agent Loop node as tools, building an "All-in-One Slipbox Agent" that can search, look up, create, edit, link, and organize a slipbox from a single natural-language request.

You'll connect the Slip Query, Slip Keyword Search, Slip Read, Slip Create, Slip Field Update, Slip Text Replace, Slip Append, Follow Slip Links, Link Slips, Slip Split Category, and Slip Merge Category nodes to an Agent Loop node as tools. Each node's detailed usage is already covered in its own tutorial, so this tutorial focuses on "how to connect many tools to an agent at once so it can combine them autonomously." Once it's built, when a user asks something like "Do we have a document on this?" or "Update this content," the agent picks and combines the right tools on its own.

Before You Begin

A slipbox must already be set up in your project before starting this tutorial. For instructions on building a slipbox, see the 🔗Slipbox Guide.

After completing this tutorial, you will be able to:

1. Connect all 11 Slipbox nodes to an Agent Loop node as tools in one workflow.

2. Use a persona and instructions to guide the agent toward selecting tools in the right order.

3. Let the agent fill in each tool node's input values at call time, instead of binding everything manually.

4. Use Chat Test and the debugger to see which tools the agent called, how many times, and in what order.

This tutorial connects the following nodes as tools. For each node's fields and standalone usage, see the linked tutorial.

• 🔗Slip Query Node

• 🔗Slip Keyword Search Node

• 🔗Slip Read Node

• 🔗Slip Create Node

• 🔗Slip Field Update Node

• 🔗Slip Text Replace Node

• 🔗Slip Append Node

• 🔗Follow Slip Links Node

• 🔗Link Slips Node

• 🔗Slip Split Category Node

• 🔗Slip Merge Category Node

• 🔗Agent Loop Node — see this document for the node's full set of basic options.

Workflow Overview

Here's what each node does:

• Agent Loop node: the core node of this workflow — it analyzes the user's request and selects and combines whichever tools are needed.

• The 11 Slipbox tool nodes: each is a dedicated tool for searching, looking up, creating, editing, linking, or organizing the slipbox. See the links under "Before You Begin" for details.

• LLM node: specifies the model the Agent Loop node and the Slip Query node use internally. The Slip Query node uses an LLM on its own to find relevant Slips, so it needs its own LLM node connection separate from the Agent Loop node's. See the 🔗LLM Node document for details.

What's Different When Connecting Multiple Tools

The Agent Loop node's full set of options (max steps, max execution time, and so on) is already covered in the 🔗Agent Loop Node document, so here we only cover the settings that matter specifically when connecting many tools at once.

There's one more important difference: how input values get bound. In the "single core node" tutorials you've seen so far, every input was bound directly to a Start Node variable via drag and drop. Tool-connected nodes work differently — an input that changes with every call, like Slip Public UUID, is left unbound. An unbound field is shown with a Parameter badge, which means "the agent fills this value in itself at call time." A value that doesn't change between runs, like Target Slipbox, is still bound at design time just as before.

An unbound input field

Step 1: Create an Agent

Create a new Agent. Enter All-in-One Slipbox Agent for Name and An agent that searches, looks up, creates, edits, and manages a slipbox for Description.

Step 2: Check the Agent Input Node

Double-click Agent Input to enter the Node Editor. Four input variables are fixed: input_message, input_files, metadata, and session_id. This tutorial uses input_message, which holds the message the user typed in chat.

Step 3: Add the Agent Loop Node and Write the Persona and Instructions

Click +Add Node to place an Agent Loop node on the canvas. In the Node Editor, drag and drop input_message onto User Input.

Click Add Variables to add Agent Persona and Agent Instructions inputs. For Agent Persona, write the agent's identity like this:

You are a specialist agent that manages the **Slipbox Master**

You are a specialist agent that manages the **Slipbox Master**

You are a specialist agent that manages the **Slipbox Master**

For Agent Instructions, write the tool selection criteria and operating order:

Finally, change Max Tools in Prompt to 11 (from the default of 10, since you're connecting exactly 11 tools).

Step 4: Add and Connect LLM Nodes

Click +Add Node to place an LLM node on the canvas, and select Anthropic-claude-4.6-sonnet for LLM Model. Connect this LLM node to the Agent Loop node's LLM Model connection point.

The Slip Query node uses an LLM internally to find relevant Slips, so add a second LLM node the same way and connect it to the Slip Query node's LLM Model connection point (you can either reuse the same LLM node or add a new one — this tutorial adds a separate LLM node with the same model). This tutorial doesn't connect Fallback LLM Model.

Step 5: Add and Connect the 11 Slipbox Tool Nodes

All 11 tool nodes follow the same pattern.

1. Click +Add Node and place a node from the Slipbox category on the canvas.

2. In the Node Editor, select Target Slipbox — this tutorial selects Slipbox Master as an example. This is the one value every tool node needs bound ahead of time.

3. Leave the remaining inputs (search term, Slip UUID, body text, and so on) unbound — the agent fills these in at call time.

4. Drag an edge from the Agent Loop node's Tools connection point to this node's In-Pin.

Repeat this pattern for the following 11 nodes.

✅ Checkpoint: Make sure all 11 tool nodes are on the canvas, each one's Target Slipbox is set, and every node is connected to the Agent Loop node's Tools. If any node's Target Slipbox is left unset, that tool will fail when the agent calls it.

Step 6: Configure the Agent Output Node

Double-click Agent Output to enter the Node Editor. Drag and drop final_answer from the Agent Loop node onto output_message.

Step 7: Run the Whole Workflow With Chat Test

Click CHAT TEST and enter a read-only question like this into the chat box:

If the agent calls Keyword Search → Slip Query → Slip Read in sequence on its own to understand the slipbox's structure and then generates an answer, it's working correctly.

Next, try a compound request that needs multiple tools at once.

With this single request, the agent combines Slip Create, Slip Append, and Link Slips as needed. Once everything is done, you'll get back a "✅ Task Complete Summary" with a table of the created Slip IDs, the section that was added, and the relationship that was linked.

Step 8: Check the Execution Process in the Debugger

The Debugger panel on the right shows the tools the agent called, in order. After each Thought step comes an Execute - [tool name] entry — if several entries appear side by side within the same step, that means those tools were called at the same time, in parallel, within one turn. For example, in the compound request above, Slip Append and Link Slips are called together in the same turn.

You can also see failed tool calls directly in the debugger. For example, if a user asks for a relationship to be created with a non-snake_case label, the Link Slips node returns error_code: INVALID_INPUT with error_message: "invalid label '...'. must be lowercase snake_case..." and fails. After seeing this failure in the debugger, the agent retries on the next turn with a corrected lowercase snake_case label, and explains what it did in its final answer.

Next Steps

🎉 Congratulations! You've successfully built the "All-in-One Slipbox Agent" workflow using Agentria.

Connect web search or other Abilities as tools alongside the Slipbox nodes, and you can extend a single agent to handle a much broader range of tasks on its own, beyond just managing a slipbox.

Agentria is a place where ideas become reality—your workflow can expand infinitely with your creativity.

• For more node examples, visit the 🔗Node Usage page.

Frequently Asked Questions

When should I use the multi-tool connection pattern?

It's a good fit when you expect users to freely request different kinds of operations — searching, creating, editing — against a single data source (here, the slipbox) in natural language. In conversational scenarios where you can't predict ahead of time which node to run in which order for every request, you can let the agent pick the right tool each time instead of building a fixed workflow.

How is connecting nodes as tools different from binding nodes directly in an Ability?

In an Ability, every input on every node has to be bound to a Start Node variable with drag and drop ahead of time. When a node is connected as a tool to the Agent Loop node instead, an input that changes with every call, like Slip Public UUID, can be left unbound — the agent fills it in itself at call time. Only values that don't change between runs, like Target Slipbox, need to be bound at design time, so there's much less to configure.

In what order does the agent call multiple tools?

The Agent Loop node analyzes the user's request to decide which tool is needed, observes the result, and repeats the cycle if another tool is needed (think → act → observe, repeated). It can also call several tools within the same turn — for a compound request like "add a section and create a linked Slip," you can see Slip Append and Link Slips called together in one turn in the debugger.

What happens if a tool call fails?

When a tool returns a failure along with an error_code, the agent uses that information on its next turn to correct the input and retry. For example, if a Korean-language label is passed to the Link Slips node and it returns INVALID_INPUT, the agent retries with a lowercase snake_case label instead, succeeds, and explains what happened in its final answer to the user.