When you select a tool to invoke, MCPFlo doesn’t ask you to hand-write a JSON payload — it generates a form directly from the tool’s declared JSON Schema, so you fill in fields the same way you’d fill out any form.

How it works

Every MCP tool declares an inputSchema describing the shape of the arguments it accepts. MCPFlo renders this schema as a live form using react-jsonschema-form (RJSF):

1. Select a tool in the capability tree.
2. Its input schema renders as a form — text fields, numbers, booleans, enums/dropdowns, and so on, matching the schema’s declared types.
3. Fill in the fields and click Invoke (or equivalent) to send the call.

Nested objects and arrays

Tools with more complex inputs are fully supported:

• Nested objects render as collapsible sub-sections matching the schema’s structure.
• Arrays render with add/remove controls, so you can build up a list of items (including arrays of objects) without editing raw JSON.

Validation

Forms validate client-side against the schema before a call goes out:

• Required fields are enforced.
• Type mismatches (e.g. text in a number field) are caught immediately.
• Constraints declared in the schema (e.g. enums, min/max, patterns) are respected.

This catches malformed input before it reaches the server, so you spend less time debugging whether a failure is your input or the server’s behavior.

When a schema is missing or minimal

If a tool declares a very loose schema (e.g. an open-ended object with no properties), the generated form will be correspondingly minimal — MCPFlo renders exactly what the schema describes, no more and no less.

After invoking

Once you invoke a tool, its result renders in the response view — see Understanding Responses — and the call is added to that tool’s history, ready to replay later (see History & Workflow).