Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.prism.byescaleira.com/llms.txt

Use this file to discover all available pages before exploring further.

​
MCP Tools

Tools are the most powerful MCP primitive. They let AI assistants execute code on your server — querying databases, calling APIs, processing files, and performing any action your application supports.

​
Registering Tools

Basic Tool
await mcp.registerTool(
    "calculate",
    description: "Perform arithmetic calculations",
    inputSchema: [
        "expression": "string"
    ],
    handler: { args in
        let expr = args["expression"] as? String ?? ""
        // Parse and evaluate (simplified)
        let result = evaluate(expr)
        return PrismMCPToolResult(
            content: [.text("Result: \(result)")],
            isError: false
        )
    }
)
Write clear, specific tool descriptions. The AI uses these to decide when and how to call your tools. A good description is the difference between a tool that gets used correctly and one that gets ignored.

​
Tool Input Schema

The inputSchema tells the AI what parameters your tool accepts:
Rich Input Schema
await mcp.registerTool(
    "search_users",
    description: "Search for users by name, email, or role. Returns matching user records.",
    inputSchema: [
        "query": "string",
        "field": "string",
        "limit": "number"
    ],
    handler: { args in
        let query = args["query"] as? String ?? ""
        let field = args["field"] as? String ?? "name"
        let limit = args["limit"] as? Int ?? 10
        
        let rows = try await db.query(
            "SELECT * FROM users WHERE \(field) LIKE ? LIMIT ?",
            parameters: [.text("%\(query)%"), .integer(limit)]
        )
        
        let results = rows.map { "- \($0["name"]): \($0["email"])" }.joined(separator: "\n")
        return PrismMCPToolResult(
            content: [.text(results.isEmpty ? "No users found" : results)],
            isError: false
        )
    }
)

​
Error Handling

When a tool encounters an error, return isError: true:
Error Result
await mcp.registerTool(
    "delete_file",
    description: "Delete a file at the given path",
    inputSchema: ["path": "string"],
    handler: { args in
        let path = args["path"] as? String ?? ""
        
        guard FileManager.default.fileExists(atPath: path) else {
            return PrismMCPToolResult(
                content: [.text("File not found: \(path)")],
                isError: true
            )
        }
        
        try FileManager.default.removeItem(atPath: path)
        return PrismMCPToolResult(
            content: [.text("Deleted: \(path)")],
            isError: false
        )
    }
)

​
Practical Examples

await mcp.registerTool(
    "query_db",
    description: "Run a read-only SQL query against the application database",
    inputSchema: ["sql": "string"],
    handler: { args in
        let sql = args["sql"] as? String ?? ""
        
        // Safety: only allow SELECT
        guard sql.trimmingCharacters(in: .whitespaces)
                  .uppercased().hasPrefix("SELECT") else {
            return PrismMCPToolResult(
                content: [.text("Only SELECT queries are allowed")],
                isError: true
            )
        }
        
        let rows = try await db.query(sql)
        let output = rows.map { row in
            row.columns.map { "\($0.key): \($0.value)" }.joined(separator: ", ")
        }.joined(separator: "\n")
        
        return PrismMCPToolResult(
            content: [.text(output.isEmpty ? "No results" : output)],
            isError: false
        )
    }
)

await mcp.registerTool(
    "http_get",
    description: "Make an HTTP GET request to a URL and return the response",
    inputSchema: ["url": "string"],
    handler: { args in
        let url = args["url"] as? String ?? ""
        let client = PrismHTTPClient()
        let response = try await client.get(url)
        let body = response.text ?? "(empty)"
        return PrismMCPToolResult(
            content: [.text("Status: \(response.statusCode)\n\n\(body)")],
            isError: response.statusCode >= 400
        )
    }
)

await mcp.registerTool(
    "find_files",
    description: "Search for files matching a pattern in the project directory",
    inputSchema: ["pattern": "string", "directory": "string"],
    handler: { args in
        let pattern = args["pattern"] as? String ?? "*"
        let dir = args["directory"] as? String ?? "."
        
        let fm = FileManager.default
        let enumerator = fm.enumerator(atPath: dir)
        var matches: [String] = []
        
        while let file = enumerator?.nextObject() as? String {
            if file.contains(pattern) {
                matches.append(file)
            }
        }
        
        return PrismMCPToolResult(
            content: [.text(matches.isEmpty ? "No files found" : matches.joined(separator: "\n"))],
            isError: false
        )
    }
)

​
Content Types

Tool results can include different content types:
Multiple Content Types
// Text content
.text("Plain text result")

// Image content (base64)
.image(data: base64String, mimeType: "image/png")

// Resource reference
.resource(uri: "file:///path/to/file", text: "File contents", mimeType: "text/plain")

​
What’s Next

Resources

Expose data as readable resources

Transports

Deploy your MCP server over stdio or HTTP