/**
* Sign-in MCP tool implementation
*/
import { CallToolRequest } from '@modelcontextprotocol/sdk/types.js';
import { SignInResponse, ERROR_CODES, ErrorResponse, MCP_TOOLS } from 'agent-timeline-shared';
import { createSession } from '../session.js';
/**
* Sign-in tool handler
* Authenticates an AI agent and starts a session
*/
export async function handleSignIn(request: CallToolRequest): Promise<SignInResponse> {
const { arguments: args } = request.params;
// Validate parameters
if (!args || typeof args !== 'object') {
throw {
error: ERROR_CODES.VALIDATION_ERROR,
message: 'Invalid arguments provided',
} as ErrorResponse;
}
const { agent_name, context } = args as { agent_name?: unknown; context?: unknown };
// Validate agent_name
if (typeof agent_name !== 'string') {
throw {
error: ERROR_CODES.VALIDATION_ERROR,
message: 'agent_name must be a string',
details: { provided: typeof agent_name },
} as ErrorResponse;
}
// Validate context (optional)
if (context !== undefined && typeof context !== 'string') {
throw {
error: ERROR_CODES.VALIDATION_ERROR,
message: 'context must be a string if provided',
details: { provided: typeof context },
} as ErrorResponse;
}
try {
// Create new session
const { sessionId, agent } = await createSession(agent_name, context);
// Return success response
const response: SignInResponse = {
session_id: sessionId,
agent_id: agent.id,
display_name: agent.display_name,
identity_key: agent.identity_key,
avatar_seed: agent.avatar_seed,
message: 'Signed in successfully',
};
return response;
} catch (error) {
// Re-throw known errors
if (typeof error === 'object' && error && 'error' in error) {
throw error;
}
// Handle unexpected errors
throw {
error: ERROR_CODES.DATABASE_ERROR,
message: `Sign-in failed: ${error instanceof Error ? error.message : 'Unknown error'}`,
} as ErrorResponse;
}
}
/**
* Tool definition for MCP server registration
*/
export const signInTool = {
name: MCP_TOOLS.SIGN_IN,
description:
'Authenticate an AI agent and start a session. Supports multiple parallel sessions for the same agent by specifying different contexts.',
inputSchema: {
type: 'object',
properties: {
agent_name: {
type: 'string',
description: 'Name of the AI agent',
minLength: 1,
maxLength: 100,
},
context: {
type: 'string',
description: 'Optional work context/task description',
maxLength: 200,
},
},
required: ['agent_name'],
},
} as const;
