M.I.M.I.R - Multi-agent Intelligent Memory & Insight Repository

/** * Orchestration Tools * * MCP tools for executing workflows with LLM agents via Mimir's orchestration API */ import { Tool } from '@modelcontextprotocol/sdk/types.js'; export const orchestrationTools: Tool[] = [ { name: 'execute_workflow', description: 'Execute a workflow with parallel LLM agents. Each task uses the copilot-api to generate actual content (code, tests, docs). Returns execution ID for tracking progress. Tasks can have dependencies and run in parallel groups.', inputSchema: { type: 'object', properties: { tasks: { type: 'array', description: 'Array of tasks to execute. Tasks with same parallelGroup run in parallel.', items: { type: 'object', properties: { id: { type: 'string', description: 'Unique task ID (e.g., "task-1")' }, title: { type: 'string', description: 'Task title' }, prompt: { type: 'string', description: 'Prompt for the LLM agent to execute' }, agentRoleDescription: { type: 'string', description: 'Role description for the agent (e.g., "TypeScript test generator")' }, recommendedModel: { type: 'string', description: 'LLM model to use (e.g., "gpt-4o", "claude-sonnet-4.5")', default: 'gpt-4o' }, parallelGroup: { type: 'number', description: 'Tasks with same group number run in parallel', default: 1 }, dependencies: { type: 'array', description: 'Array of task IDs this task depends on', items: { type: 'string' }, default: [] }, successCriteria: { type: 'array', description: 'Criteria for task success', items: { type: 'string' }, default: [] }, maxRetries: { type: 'number', description: 'Maximum retry attempts', default: 2 } }, required: ['id', 'title', 'prompt', 'agentRoleDescription'] } } }, required: ['tasks'] } }, { name: 'get_execution_status', description: 'Get the current status of a workflow execution', inputSchema: { type: 'object', properties: { execution_id: { type: 'string', description: 'ID of the execution (returned from execute_workflow)' } }, required: ['execution_id'] } }, { name: 'get_execution_results', description: 'Get the results from a completed workflow execution, including all task outputs and deliverables', inputSchema: { type: 'object', properties: { execution_id: { type: 'string', description: 'ID of the execution' } }, required: ['execution_id'] } }, { name: 'cancel_execution', description: 'Cancel a running workflow execution', inputSchema: { type: 'object', properties: { execution_id: { type: 'string', description: 'ID of the execution to cancel' } }, required: ['execution_id'] } } ];