tuningengines-cli 0.3.5 → 0.4.0

package/dist/mcp.js

@@ -7,13 +7,22 @@ const types_js_1 = require("@modelcontextprotocol/sdk/types.js");
 const client_1 = require("./client");
 const config_1 = require("./config");
 async function startMcpServer() {
-    const client = new client_1.TuningEnginesClient({
-        apiKey: (0, config_1.getApiKey)(),
-        apiUrl: (0, config_1.getApiUrl)(),
-    });
-    const server = new index_js_1.Server({ name: "tuning-engines", version: "0.3.5" }, {
+    // Lazy client initialization — deferred until a tool is called.
+    // This allows the server to start and list tools without a valid API key,
+    // which is required for Glama inspection and tool detection.
+    let _client = null;
+    const getClient = () => {
+        if (!_client) {
+            _client = new client_1.TuningEnginesClient({
+                apiKey: (0, config_1.getApiKey)(),
+                apiUrl: (0, config_1.getApiUrl)(),
+            });
+        }
+        return _client;
+    };
+    const server = new index_js_1.Server({ name: "tuning-engines", version: "0.3.6" }, {
         capabilities: { tools: {} },
-        instructions: "Tuning Engines — Domain-specific fine-tuning of open-source LLMs and SLMs. Own your sovereign model with zero infrastructure.\n\n" +
+        instructions: "Tuning Engines — Domain-specific fine-tuning of open-source LLMs and SLMs, plus a Marketplace of pre-built models and datasets. Own your sovereign model with zero infrastructure.\n\n" +
             "USE THIS SERVER WHEN the user wants to:\n" +
             "- Fine-tune, train, or customize any open-source LLM/SLM on their data\n" +
             "- Build a sovereign AI model trained on their organization's code, documents, or domain data\n" +
@@ -21,11 +30,17 @@ async function startMcpServer() {
             "- Train using LoRA, QLoRA, or full fine-tuning methods\n" +
             "- Estimate the cost of fine-tuning a model\n" +
             "- Check training job status, manage trained models, or check billing\n" +
-            "- Export or import models to/from S3\n\n" +
+            "- Export or import models to/from S3\n" +
+            "- Browse and export pre-built models and datasets from the Marketplace\n\n" +
+            "MARKETPLACE:\n" +
+            "Pre-built, ready-to-use fine-tuned models and datasets curated by the platform. " +
+            "Browse the catalog, view details, and export directly to your S3 bucket. " +
+            "Credits are charged per export based on the item's price.\n" +
+            "Workflow: list_catalog_models → get_catalog_model → export_catalog_model → catalog_export_status\n\n" +
             "SPECIALIZED TUNING AGENTS (more coming):\n" +
             "- Cody (code_repo): Code autocomplete and inline suggestions via QLoRA/Axolotl\n" +
             "- SIERA (sera_code_repo): Bug-fix and error resolution via AllenAI Open Coding Agents\n\n" +
-            "TYPICAL WORKFLOW: estimate_job → create_job → job_status (poll until done) → list_models\n\n" +
+            "TYPICAL TRAINING WORKFLOW: estimate_job → create_job → job_status (poll until done) → list_models\n\n" +
             "Supports 1B to 72B parameter models from Qwen, Llama, DeepSeek, Mistral, Gemma, Phi, StarCoder, and CodeLlama families.\n" +
             "Zero infrastructure — GPU provisioning, training orchestration, and model delivery fully managed.",
     });
@@ -75,9 +90,10 @@ async function startMcpServer() {
                     "Supports quality_tier='low' (faster) or quality_tier='high' (deeper analysis, more training data).\n\n" +
                     "SUPPORTED BASE MODELS (by size):\n" +
                     "- 3B: Qwen/Qwen2.5-Coder-3B-Instruct\n" +
-                    "- 7B: codellama/CodeLlama-7b-hf, deepseek-ai/deepseek-coder-7b-instruct-v1.5, Qwen/Qwen2.5-Coder-7B-Instruct\n" +
-                    "- 13-15B: codellama/CodeLlama-13b-Instruct-hf, bigcode/starcoder2-15b, Qwen/Qwen2.5-Coder-14B-Instruct\n" +
-                    "- 32-34B: deepseek-ai/deepseek-coder-33b-instruct, codellama/CodeLlama-34b-Instruct-hf, Qwen/Qwen2.5-Coder-32B-Instruct\n" +
+                    "- 7-8B: codellama/CodeLlama-7b-hf, deepseek-ai/deepseek-coder-7b-instruct-v1.5, Qwen/Qwen2.5-Coder-7B-Instruct, Qwen/Qwen3-8B\n" +
+                    "- 13-15B: codellama/CodeLlama-13b-Instruct-hf, bigcode/starcoder2-15b, Qwen/Qwen2.5-Coder-14B-Instruct, Qwen/Qwen3-14B\n" +
+                    "- 22-27B: mistralai/Codestral-22B-v0.1, google/gemma-2-27b\n" +
+                    "- 30-34B: deepseek-ai/deepseek-coder-33b-instruct, codellama/CodeLlama-34b-Instruct-hf, Qwen/Qwen2.5-Coder-32B-Instruct, Qwen/Qwen3-Coder-30B-A3B, Qwen/Qwen3-32B\n" +
                     "- 70-72B: codellama/CodeLlama-70b-Instruct-hf, meta-llama/Llama-3.1-70B-Instruct, Qwen/Qwen2.5-72B-Instruct\n\n" +
                     "TYPICAL WORKFLOW: estimate_job first to check cost, then create_job, then job_status to monitor progress.",
                 inputSchema: {
@@ -167,7 +183,10 @@ async function startMcpServer() {
             },
             {
                 name: "retry_job",
-                description: "Retry a failed fine-tuning job from its last checkpoint. Creates a new job that resumes training where the failed one stopped, saving GPU time. Each retry is billed separately.",
+                description: "Retry a failed fine-tuning job from its last checkpoint. Creates a new job that resumes training where the failed one stopped, saving GPU time. Each retry is billed separately.\n\n" +
+                    "IMPORTANT: This tool fetches a cost estimate and includes it in the response. " +
+                    "You MUST show the estimate to the user and get their explicit approval before considering the retry confirmed. " +
+                    "The retry is submitted automatically (the server validates balance), but always present the cost to the user.",
                 inputSchema: {
                     type: "object",
                     properties: {
@@ -203,6 +222,10 @@ async function startMcpServer() {
                             type: "number",
                             description: "Approximate repository size in MB (helps refine the estimate)",
                         },
+                        use_case: {
+                            type: "string",
+                            description: "Agent to use for the estimate (e.g. 'code_repo' for Cody, 'sera_code_repo' for SIERA). Defaults to code_repo.",
+                        },
                     },
                 },
             },
@@ -268,10 +291,15 @@ async function startMcpServer() {
             },
             {
                 name: "list_supported_models",
-                description: "List the supported base HuggingFace models available for fine-tuning on Tuning Engines.",
+                description: "List the supported base HuggingFace models available for fine-tuning on Tuning Engines. Optionally filter by agent to see only compatible models.",
                 inputSchema: {
                     type: "object",
-                    properties: {},
+                    properties: {
+                        agent: {
+                            type: "string",
+                            description: "Filter models compatible with this agent (e.g. 'code_repo', 'sera_code_repo'). Omit to see all models.",
+                        },
+                    },
                 },
             },
             {
@@ -330,6 +358,285 @@ async function startMcpServer() {
                     required: ["model_id"],
                 },
             },
+            {
+                name: "list_catalog_models",
+                description: "List available pre-built models and datasets from the Tuning Engines Marketplace. " +
+                    "These are platform-owned, ready-to-use assets that can be exported to your S3 bucket. " +
+                    "Returns name, description, base model, size, export price, and category.",
+                inputSchema: {
+                    type: "object",
+                    properties: {
+                        category: {
+                            type: "string",
+                            description: "Filter by category (e.g. 'code', 'bug-fix', 'general'). Omit to see all.",
+                        },
+                    },
+                },
+            },
+            {
+                name: "get_catalog_model",
+                description: "Get detailed information about a specific pre-built model or dataset from the Marketplace including description, pricing, and export options.",
+                inputSchema: {
+                    type: "object",
+                    properties: {
+                        model_id: { type: "string", description: "Catalog model ID (UUID)" },
+                    },
+                    required: ["model_id"],
+                },
+            },
+            {
+                name: "export_catalog_model",
+                description: "Export a pre-built model or dataset from the Marketplace to your S3 bucket. " +
+                    "Credits will be charged based on the export price upon successful completion.",
+                inputSchema: {
+                    type: "object",
+                    properties: {
+                        model_id: { type: "string", description: "Catalog model ID (UUID) to export" },
+                        s3_bucket: { type: "string", description: "Destination S3 bucket name" },
+                        s3_prefix: {
+                            type: "string",
+                            description: "Optional S3 key prefix for the exported model",
+                        },
+                        s3_access_key_id: { type: "string", description: "AWS access key ID" },
+                        s3_secret_access_key: { type: "string", description: "AWS secret access key" },
+                        s3_region: { type: "string", description: "AWS region (e.g. us-east-1)" },
+                    },
+                    required: ["model_id", "s3_bucket", "s3_access_key_id", "s3_secret_access_key", "s3_region"],
+                },
+            },
+            {
+                name: "catalog_export_status",
+                description: "Check the status of a Marketplace export operation. Returns status, charge info, and any error messages.",
+                inputSchema: {
+                    type: "object",
+                    properties: {
+                        model_id: { type: "string", description: "Catalog model ID (UUID)" },
+                        export_id: { type: "string", description: "Export operation ID (UUID)" },
+                    },
+                    required: ["model_id", "export_id"],
+                },
+            },
+            // --- Datasets ---
+            {
+                name: "list_datasets",
+                description: "List datasets available for training and evaluation. Datasets can be uploaded from S3 and used for fine-tuning or model evaluation.",
+                inputSchema: {
+                    type: "object",
+                    properties: {
+                        limit: { type: "number", description: "Max results (default 20)" },
+                    },
+                },
+            },
+            {
+                name: "show_dataset",
+                description: "Get details of a specific dataset including status, source, and metadata.",
+                inputSchema: {
+                    type: "object",
+                    properties: {
+                        dataset_id: { type: "string", description: "Dataset ID (UUID)" },
+                    },
+                    required: ["dataset_id"],
+                },
+            },
+            {
+                name: "create_dataset",
+                description: "Create a new dataset by importing from S3. Datasets can be used for fine-tuning or model evaluation.",
+                inputSchema: {
+                    type: "object",
+                    properties: {
+                        name: { type: "string", description: "Name for the dataset" },
+                        description: { type: "string", description: "Description of the dataset contents" },
+                        source_type: { type: "string", description: "Source type (e.g. 's3')" },
+                        s3_url: { type: "string", description: "S3 URL of the dataset (e.g. s3://bucket/path/data.jsonl)" },
+                        s3_access_key_id: { type: "string", description: "AWS access key ID" },
+                        s3_secret_access_key: { type: "string", description: "AWS secret access key" },
+                        s3_region: { type: "string", description: "AWS region (e.g. us-east-1)" },
+                        for_evaluation: { type: "boolean", description: "Whether this dataset is for evaluation (default: false)" },
+                    },
+                    required: ["name", "source_type"],
+                },
+            },
+            {
+                name: "delete_dataset",
+                description: "Delete a dataset from the platform.",
+                inputSchema: {
+                    type: "object",
+                    properties: {
+                        dataset_id: { type: "string", description: "Dataset ID to delete" },
+                    },
+                    required: ["dataset_id"],
+                },
+            },
+            {
+                name: "dataset_status",
+                description: "Check the status of a dataset import or processing operation.",
+                inputSchema: {
+                    type: "object",
+                    properties: {
+                        dataset_id: { type: "string", description: "Dataset ID (UUID)" },
+                    },
+                    required: ["dataset_id"],
+                },
+            },
+            // --- Evaluations ---
+            {
+                name: "list_evaluations",
+                description: "List model evaluations. Evaluations run your trained models against benchmark datasets using various evaluators to measure quality.",
+                inputSchema: {
+                    type: "object",
+                    properties: {
+                        status: {
+                            type: "string",
+                            description: "Filter by status: queued, running, succeeded, failed, canceled",
+                        },
+                        limit: { type: "number", description: "Max results (default 20)" },
+                    },
+                },
+            },
+            {
+                name: "show_evaluation",
+                description: "Get full details of a specific evaluation including status, scores, metrics, and comparison data.",
+                inputSchema: {
+                    type: "object",
+                    properties: {
+                        evaluation_id: { type: "string", description: "Evaluation ID (UUID)" },
+                    },
+                    required: ["evaluation_id"],
+                },
+            },
+            {
+                name: "create_evaluation",
+                description: "Create a new model evaluation. Run your trained model or a base model against a dataset using selected evaluators. " +
+                    "Use list_evaluators to see available evaluators (e.g. code_execution, similarity, llm_judge).",
+                inputSchema: {
+                    type: "object",
+                    properties: {
+                        name: { type: "string", description: "Name for this evaluation run" },
+                        user_model_id: {
+                            type: "string",
+                            description: "ID of your trained model to evaluate. Either this or base_model is required.",
+                        },
+                        base_model: {
+                            type: "string",
+                            description: "HuggingFace model ID to evaluate (e.g. 'Qwen/Qwen2.5-Coder-7B-Instruct'). Either this or user_model_id is required.",
+                        },
+                        dataset_id: {
+                            type: "string",
+                            description: "ID of the evaluation dataset to use. Must be a dataset marked for_evaluation.",
+                        },
+                        evaluator_ids: {
+                            type: "array",
+                            items: { type: "string" },
+                            description: "List of evaluator IDs to run (use list_evaluators to see options)",
+                        },
+                        max_samples: {
+                            type: "number",
+                            description: "Maximum samples to evaluate (default: all)",
+                        },
+                    },
+                    required: ["dataset_id", "evaluator_ids"],
+                },
+            },
+            {
+                name: "cancel_evaluation",
+                description: "Cancel a running or queued evaluation.",
+                inputSchema: {
+                    type: "object",
+                    properties: {
+                        evaluation_id: { type: "string", description: "Evaluation ID to cancel" },
+                    },
+                    required: ["evaluation_id"],
+                },
+            },
+            {
+                name: "evaluation_status",
+                description: "Get live status of an evaluation including progress and current metrics.",
+                inputSchema: {
+                    type: "object",
+                    properties: {
+                        evaluation_id: { type: "string", description: "Evaluation ID (UUID)" },
+                    },
+                    required: ["evaluation_id"],
+                },
+            },
+            {
+                name: "list_evaluators",
+                description: "List available evaluators for model evaluation. Evaluators measure different aspects of model quality like code execution, similarity, or LLM-based judgment.",
+                inputSchema: {
+                    type: "object",
+                    properties: {},
+                },
+            },
+            {
+                name: "estimate_evaluation",
+                description: "Get a cost estimate for an evaluation before running it.",
+                inputSchema: {
+                    type: "object",
+                    properties: {
+                        user_model_id: { type: "string", description: "ID of your trained model" },
+                        base_model: { type: "string", description: "Or a HuggingFace model ID" },
+                        dataset_id: { type: "string", description: "Evaluation dataset ID" },
+                        evaluator_ids: {
+                            type: "array",
+                            items: { type: "string" },
+                            description: "List of evaluator IDs",
+                        },
+                        max_samples: { type: "number", description: "Max samples to evaluate" },
+                    },
+                    required: ["dataset_id", "evaluator_ids"],
+                },
+            },
+            // --- Inference ---
+            {
+                name: "list_inference_models",
+                description: "List models available for inference through the Tuning Engines inference API. " +
+                    "Includes both platform models and your deployed trained models.",
+                inputSchema: {
+                    type: "object",
+                    properties: {},
+                },
+            },
+            {
+                name: "inference_usage",
+                description: "Get inference API usage statistics including request counts, token usage, and costs.",
+                inputSchema: {
+                    type: "object",
+                    properties: {
+                        start_date: { type: "string", description: "Start date (YYYY-MM-DD)" },
+                        end_date: { type: "string", description: "End date (YYYY-MM-DD)" },
+                        model: { type: "string", description: "Filter by model name" },
+                    },
+                },
+            },
+            {
+                name: "get_inference_jwt",
+                description: "Get a JWT token for authenticating with the Tuning Engines inference API. " +
+                    "Use this to make direct API calls to the inference endpoint.",
+                inputSchema: {
+                    type: "object",
+                    properties: {},
+                },
+            },
+            // --- Agents ---
+            {
+                name: "list_agents",
+                description: "List available agents configured for your organization. Agents are AI assistants with specific capabilities and tool access.",
+                inputSchema: {
+                    type: "object",
+                    properties: {},
+                },
+            },
+            {
+                name: "show_agent",
+                description: "Get details of a specific agent including capabilities, tools, and configuration.",
+                inputSchema: {
+                    type: "object",
+                    properties: {
+                        agent_id: { type: "string", description: "Agent ID" },
+                    },
+                    required: ["agent_id"],
+                },
+            },
         ],
     }));
     // Handle tool calls
@@ -339,13 +646,13 @@ async function startMcpServer() {
             let result;
             switch (name) {
                 case "list_jobs":
-                    result = await client.listJobs({
+                    result = await getClient().listJobs({
                         status: args?.status,
                         limit: args?.limit,
                     });
                     break;
                 case "show_job":
-                    result = await client.getJob(args.job_id);
+                    result = await getClient().getJob(args.job_id);
                     break;
                 case "create_job":
                     if (!args?.base_model && !args?.base_user_model_id) {
@@ -354,7 +661,7 @@ async function startMcpServer() {
                             isError: true,
                         };
                     }
-                    result = await client.createJob({
+                    result = await getClient().createJob({
                         base_model: args?.base_model,
                         base_user_model_id: args?.base_user_model_id,
                         output_name: args.output_name,
@@ -371,14 +678,34 @@ async function startMcpServer() {
                     });
                     break;
                 case "cancel_job":
-                    result = await client.cancelJob(args.job_id);
+                    result = await getClient().cancelJob(args.job_id);
                     break;
                 case "job_status":
-                    result = await client.getJobStatus(args.job_id);
+                    result = await getClient().getJobStatus(args.job_id);
                     break;
-                case "retry_job":
-                    result = await client.retryJob(args.job_id, args?.github_token);
+                case "retry_job": {
+                    // Fetch job details and estimate before retrying so the AI can show cost
+                    const retryJobId = args.job_id;
+                    const jobDetails = await getClient().getJob(retryJobId);
+                    let retryEstimate = null;
+                    try {
+                        retryEstimate = await getClient().estimateJob({
+                            base_model: jobDetails.base_model,
+                            num_epochs: jobDetails.num_epochs,
+                            max_examples: jobDetails.max_examples,
+                            use_case: jobDetails.agent,
+                        });
+                    }
+                    catch (estErr) {
+                        // Estimate failed — continue with retry (server validates balance)
+                    }
+                    const retryResult = await getClient().retryJob(retryJobId, args?.github_token);
+                    result = {
+                        ...retryResult,
+                        retry_estimate: retryEstimate,
+                    };
                     break;
+                }
                 case "estimate_job":
                     if (!args?.base_model && !args?.base_user_model_id) {
                         return {
@@ -386,16 +713,17 @@ async function startMcpServer() {
                             isError: true,
                         };
                     }
-                    result = await client.estimateJob({
+                    result = await getClient().estimateJob({
                         base_model: args?.base_model,
                         base_user_model_id: args?.base_user_model_id,
                         num_epochs: args?.num_epochs,
                         max_examples: args?.max_examples,
                         repo_size_mb: args?.repo_size_mb,
+                        use_case: args?.use_case,
                     });
                     break;
                 case "validate_s3":
-                    result = await client.validateS3({
+                    result = await getClient().validateS3({
                         s3_bucket: args.s3_bucket,
                         s3_access_key_id: args.s3_access_key_id,
                         s3_secret_access_key: args.s3_secret_access_key,
@@ -403,25 +731,25 @@ async function startMcpServer() {
                     });
                     break;
                 case "list_models":
-                    result = await client.listUserModels();
+                    result = await getClient().listUserModels();
                     break;
                 case "show_model":
-                    result = await client.getUserModel(args.model_id);
+                    result = await getClient().getUserModel(args.model_id);
                     break;
                 case "delete_model":
-                    result = await client.deleteUserModel(args.model_id);
+                    result = await getClient().deleteUserModel(args.model_id);
                     break;
                 case "get_balance":
-                    result = await client.getBilling();
+                    result = await getClient().getBilling();
                     break;
                 case "get_account":
-                    result = await client.getAccount();
+                    result = await getClient().getAccount();
                     break;
                 case "list_supported_models":
-                    result = await client.listModels();
+                    result = await getClient().listModels({ agent: args?.agent });
                     break;
                 case "import_model":
-                    result = await client.importModel({
+                    result = await getClient().importModel({
                         name: args.name,
                         source_s3_url: args.source_s3_url,
                         base_model: args.base_model,
@@ -431,7 +759,7 @@ async function startMcpServer() {
                     });
                     break;
                 case "export_model":
-                    result = await client.exportModel(args.model_id, {
+                    result = await getClient().exportModel(args.model_id, {
                         s3_bucket: args.s3_bucket,
                         s3_prefix: args?.s3_prefix,
                         s3_access_key_id: args.s3_access_key_id,
@@ -441,7 +769,119 @@ async function startMcpServer() {
                     });
                     break;
                 case "model_status":
-                    result = await client.getUserModelStatus(args.model_id);
+                    result = await getClient().getUserModelStatus(args.model_id);
+                    break;
+                case "list_catalog_models":
+                    result = await getClient().listCatalogModels({
+                        category: args?.category,
+                    });
+                    break;
+                case "get_catalog_model":
+                    result = await getClient().getCatalogModel(args.model_id);
+                    break;
+                case "export_catalog_model":
+                    result = await getClient().exportCatalogModel(args.model_id, {
+                        s3_bucket: args.s3_bucket,
+                        s3_prefix: args?.s3_prefix,
+                        s3_access_key_id: args.s3_access_key_id,
+                        s3_secret_access_key: args.s3_secret_access_key,
+                        s3_region: args.s3_region,
+                    });
+                    break;
+                case "catalog_export_status":
+                    result = await getClient().getCatalogExportStatus(args.model_id, args.export_id);
+                    break;
+                // --- Datasets ---
+                case "list_datasets":
+                    result = await getClient().listDatasets({
+                        limit: args?.limit,
+                    });
+                    break;
+                case "show_dataset":
+                    result = await getClient().getDataset(args.dataset_id);
+                    break;
+                case "create_dataset":
+                    result = await getClient().createDataset({
+                        name: args.name,
+                        description: args?.description,
+                        source_type: args.source_type,
+                        s3_url: args?.s3_url,
+                        s3_access_key_id: args?.s3_access_key_id,
+                        s3_secret_access_key: args?.s3_secret_access_key,
+                        s3_region: args?.s3_region,
+                        for_evaluation: args?.for_evaluation,
+                    });
+                    break;
+                case "delete_dataset":
+                    result = await getClient().deleteDataset(args.dataset_id);
+                    break;
+                case "dataset_status":
+                    result = await getClient().getDatasetStatus(args.dataset_id);
+                    break;
+                // --- Evaluations ---
+                case "list_evaluations":
+                    result = await getClient().listEvaluations({
+                        status: args?.status,
+                        limit: args?.limit,
+                    });
+                    break;
+                case "show_evaluation":
+                    result = await getClient().getEvaluation(args.evaluation_id);
+                    break;
+                case "create_evaluation":
+                    if (!args?.user_model_id && !args?.base_model) {
+                        return {
+                            content: [{ type: "text", text: "Error: either user_model_id or base_model is required" }],
+                            isError: true,
+                        };
+                    }
+                    result = await getClient().createEvaluation({
+                        name: args?.name,
+                        user_model_id: args?.user_model_id,
+                        base_model: args?.base_model,
+                        dataset_id: args.dataset_id,
+                        evaluator_ids: args.evaluator_ids,
+                        max_samples: args?.max_samples,
+                    });
+                    break;
+                case "cancel_evaluation":
+                    result = await getClient().cancelEvaluation(args.evaluation_id);
+                    break;
+                case "evaluation_status":
+                    result = await getClient().getEvaluationStatus(args.evaluation_id);
+                    break;
+                case "list_evaluators":
+                    result = await getClient().listEvaluators();
+                    break;
+                case "estimate_evaluation":
+                    result = await getClient().estimateEvaluation({
+                        user_model_id: args?.user_model_id,
+                        base_model: args?.base_model,
+                        dataset_id: args.dataset_id,
+                        evaluator_ids: args.evaluator_ids,
+                        max_samples: args?.max_samples,
+                    });
+                    break;
+                // --- Inference ---
+                case "list_inference_models":
+                    result = await getClient().listInferenceModels();
+                    break;
+                case "inference_usage":
+                    result = await getClient().getInferenceUsage({
+                        start_date: args?.start_date,
+                        end_date: args?.end_date,
+                        model: args?.model,
+                    });
+                    break;
+                case "get_inference_jwt":
+                    result = await getClient().getInferenceJwt();
+                    break;
+                // --- Agents ---
+                case "list_agents":
+                    result = await getClient().listAgents();
+                    break;
+                case "show_agent":
+                    result = await getClient().getAgent(args.agent_id);
                     break;
                 default:
                     return {