Step 4 of 5

Tool config

Name your tool, write the LLM-facing description, and define the input/output schemas. The description is the primary signal the model uses when deciding whether to call this tool.

Tool identity
Machine-readable name and LLM-facing description
Machine-readable identifier used in MCP tool calls. Only lowercase letters, numbers and underscores. Cannot be changed after publishing.
Human-readable name shown in the console and audit log.
Short imperative sentence shown in the MCP tool manifest.
LLM-facing description. Be precise — this is the primary signal the model uses to decide whether to invoke this tool.
Writing effective descriptions
Start with a verb ("Creates", "Fetches", "Updates"). Mention key input requirements. Describe what a successful response looks like. Avoid vague phrasing like "handles" or "manages". See description best practices →
Input schema
Parameters the LLM passes when calling this tool
JSON Schema draft-07
json / input-schema
Output schema
Shape of the data returned to the LLM after a successful call
JSON Schema draft-07
json / output-schema
Schemas are protocol-agnostic
These schemas describe the MCP contract, not the upstream wire format. Field mapping in Step 5 translates between the two. You can upload a sample JSON response to auto-generate the output schema.