Tools Agent
Tools Agent
Section titled “Tools Agent”Generate an action plan and execute it. Can use external tools.
What it does
Section titled “What it does”Creates an AI agent that can use external tools to complete complex tasks by planning and executing actions.
When to use it
Section titled “When to use it”Use this agent when you need an LLM to interact with external APIs, databases, or services to accomplish multi-step tasks.
Tools that take actions
Section titled “Tools that take actions”Connect tool dependency nodes into this agent’s tools input to give it capabilities. Some tools only read or compute; others take real actions on your behalf. The action tools are gated — the model can call them, but each call pauses for your approval before it runs:
- Browser Tool — click, fill, type, select, submit, or navigate the active page.
- HTTP Request Tool — send HTTP requests to APIs.
- MCP Client Tool — call the tools of a remote MCP server.
You approve or reject each action in the Executions pane. This is automatic and cannot be turned off — see Action Approval for why.
Inputs and settings
Section titled “Inputs and settings”| Setting | Notes |
|---|---|
| System message | Source-backed field from the node schema. |
| Options | Source-backed field from the node schema. |
Outputs
Section titled “Outputs”Returns node-specific output described by the implementation and visible in workflow execution data.
Dependencies and credentials
Section titled “Dependencies and credentials”- No explicit credential or node dependency is declared in the node description.
Memory
Section titled “Memory”Connect a chat memory to the agent’s memory input to give it a conversation:
- Before calling the model, the agent loads the conversation’s earlier messages and adds them to the prompt as chat history.
- It saves the incoming prompt as a
usermessage. - After the model answers, it saves the reply as an
assistantmessage (structured replies from an output parser are saved as JSON text).
With Local Memory or Persistent Chat Memory the conversation is kept across runs; Window Buffer Memory only lasts for one run.
This is also how a Chat Trigger gets its replies: connect the same memory to the trigger and to this agent, and set the prompt to {{ $json.chatInput }}. The chat window shows the messages this agent saves.
Example workflow
Section titled “Example workflow”Connect Tools Agent to an AI agent or dependency input that accepts this dependency type, then run the agent with data from previous nodes.
Troubleshooting
Section titled “Troubleshooting”- “This agent requires that the bind_tools() method be implemented on the input model” — the connected chat model can’t call tools. Cloud models support it, and so do the on-device Transformers Chat and Web LLM models (prompt-based tool calling). Check that the model node is up to date and connected to the agent’s Model input.
- Check that required settings are present before running the node.
- If the node uses browser page data, run it on the target tab after the page has loaded.
- If it calls an external service, verify credentials, permissions, and rate limits.
- This node has source tests; use them as the reference for edge-case behavior during maintenance.