Core
Tool
Learn what an MCP tool is, how tools/list exposes tool schemas, and how Astrail turns API operations into safe hosted MCP tools.
Definition
A tool is an MCP server capability that a model can invoke to interact with an external system. In practice, tools wrap API calls, computations, search operations, workflow actions, or other bounded operations with a machine-readable name, description, and input schema.
How Astrail Uses It
Astrail generates tools from OpenAPI operations, website read actions, and owned SDK endpoints. Each generated tool is routed through endpoint maps, auth policy, input validation, rate limits, and runtime logging before it reaches an upstream system.
Example
{
"name": "list_pets",
"description": "List pets available in the store.",
"inputSchema": {
"type": "object",
"properties": { "limit": { "type": "integer" } }
}
}Implementation Checklist
Give every tool a stable name that describes the action, not the implementation.
Keep descriptions concrete enough for an agent to decide when to use the tool.
Validate tool input before dispatching to an upstream API.
Separate public tools from private or credential-gated tools.
FAQ
How is an MCP tool different from a normal API endpoint?
An API endpoint is designed for application code. An MCP tool adds agent-facing metadata, schemas, and invocation semantics so a model can discover and request the operation through a standard protocol.
Should every API operation become a separate MCP tool?
Not always. Large APIs often work better with fewer high-signal tools, docs search, or Code Mode execution so the model is not flooded with hundreds of low-value choices.