Skip to main content
Glama

relay_run

Execute a single AI model call to test prompts before building full workflows. Returns output, token usage, estimated provider cost, and trace URL.

Instructions

Execute a single AI model call. Useful for testing prompts before building full workflows. Returns output, token usage, estimated provider cost, and trace URL. Note: Cost tracks your provider bill (OpenAI/Anthropic), not RelayPlane fees - we're BYOK.

Input Schema

TableJSON Schema
NameRequiredDescriptionDefault
modelYesModel in provider:model format (e.g., 'openai:gpt-4o', 'anthropic:claude-3-5-sonnet-20241022')
promptYesThe user prompt to send
systemPromptNoOptional system prompt
schemaNoOptional JSON schema for structured output

Implementation Reference

  • Main tool handler that orchestrates model execution: parses model, checks budget and config, calls provider-specific execute functions, calculates costs, stores run data, handles errors.
    export async function relayRun(input: RelayRunInput): Promise<RelayRunResponse> {
      const startTime = Date.now();
      const runId = generateRunId();
      const config = getConfig();
    
      try {
        // Parse model
        const { provider, modelId } = parseModel(input.model);
    
        // Check provider is configured
        if (!isProviderConfigured(provider)) {
          throw new Error(
            `Provider "${provider}" is not configured. Set ${provider.toUpperCase()}_API_KEY environment variable.`
          );
        }
    
        // Estimate cost and check budget
        const estimatedCost = estimateProviderCost(input.model, input.prompt, input.systemPrompt);
        const budgetCheck = checkBudget(estimatedCost);
    
        if (!budgetCheck.allowed) {
          throw new Error(budgetCheck.error);
        }
    
        // Get API key
        const apiKey = getProviderKey(provider)!;
    
        // Execute based on provider
        let result: { output: any; promptTokens: number; completionTokens: number };
    
        switch (provider) {
          case 'openai':
            result = await executeOpenAI(apiKey, modelId, input.prompt, input.systemPrompt, input.schema);
            break;
          case 'anthropic':
            result = await executeAnthropic(apiKey, modelId, input.prompt, input.systemPrompt, input.schema);
            break;
          case 'google':
            result = await executeGoogle(apiKey, modelId, input.prompt, input.systemPrompt);
            break;
          case 'xai':
            result = await executeXAI(apiKey, modelId, input.prompt, input.systemPrompt);
            break;
          default:
            throw new Error(`Unsupported provider: ${provider}`);
        }
    
        const durationMs = Date.now() - startTime;
        const actualCost = calculateActualCost(input.model, result.promptTokens, result.completionTokens);
    
        // Record actual cost
        recordCost(actualCost);
    
        const response: RelayRunResponse = {
          success: true,
          output: result.output,
          model: input.model,
          usage: {
            promptTokens: result.promptTokens,
            completionTokens: result.completionTokens,
            totalTokens: result.promptTokens + result.completionTokens,
            estimatedProviderCostUsd: actualCost,
          },
          durationMs,
          runId,
          traceUrl: `${config.traceUrlBase}/${runId}`,
        };
    
        // Store run
        addRun({
          runId,
          type: 'single',
          model: input.model,
          success: true,
          startTime: new Date(startTime),
          endTime: new Date(),
          durationMs,
          usage: response.usage,
          input: { prompt: input.prompt, systemPrompt: input.systemPrompt },
          output: result.output,
        });
    
        return response;
      } catch (error) {
        const durationMs = Date.now() - startTime;
        const errorMessage = error instanceof Error ? error.message : String(error);
    
        const response: RelayRunResponse = {
          success: false,
          output: '',
          model: input.model,
          usage: {
            promptTokens: 0,
            completionTokens: 0,
            totalTokens: 0,
            estimatedProviderCostUsd: 0,
          },
          durationMs,
          runId,
          traceUrl: `${config.traceUrlBase}/${runId}`,
          error: {
            code: 'EXECUTION_ERROR',
            message: errorMessage,
          },
        };
    
        // Store failed run
        addRun({
          runId,
          type: 'single',
          model: input.model,
          success: false,
          startTime: new Date(startTime),
          endTime: new Date(),
          durationMs,
          usage: response.usage,
          input: { prompt: input.prompt, systemPrompt: input.systemPrompt },
          error: errorMessage,
        });
    
        return response;
      }
  • Zod schema for validating input to the relay_run tool.
    export const relayRunSchema = z.object({
      model: z
        .string()
        .describe("Model in provider:model format (e.g., 'openai:gpt-4o', 'anthropic:claude-3-5-sonnet-20241022')"),
      prompt: z.string().describe('The user prompt to send'),
      systemPrompt: z.string().optional().describe('Optional system prompt'),
      schema: z.object({}).passthrough().optional().describe('Optional JSON schema for structured output'),
    });
  • Tool definition object used for MCP registration, including name, description, and input schema.
    export const relayRunDefinition = {
      name: 'relay_run',
      description:
        "Execute a single AI model call. Useful for testing prompts before building full workflows. Returns output, token usage, estimated provider cost, and trace URL. Note: Cost tracks your provider bill (OpenAI/Anthropic), not RelayPlane fees - we're BYOK.",
      inputSchema: {
        type: 'object' as const,
        properties: {
          model: {
            type: 'string',
            description: "Model in provider:model format (e.g., 'openai:gpt-4o', 'anthropic:claude-3-5-sonnet-20241022')",
          },
          prompt: {
            type: 'string',
            description: 'The user prompt to send',
          },
          systemPrompt: {
            type: 'string',
            description: 'Optional system prompt',
          },
          schema: {
            type: 'object',
            description: 'Optional JSON schema for structured output',
          },
        },
        required: ['model', 'prompt'],
      },
    };
  • src/server.ts:59-67 (registration)
    Array of all tool definitions registered with the MCP server for the listTools request.
    const TOOLS = [
      relayModelsListDefinition,
      relayRunDefinition,
      relayWorkflowRunDefinition,
      relayWorkflowValidateDefinition,
      relaySkillsListDefinition,
      relayRunsListDefinition,
      relayRunGetDefinition,
    ];
  • src/server.ts:114-118 (registration)
    Dispatch case in MCP callTool handler that validates input with schema and invokes the relayRun handler function.
    case 'relay_run': {
      const parsed = relayRunSchema.parse(args);
      result = await relayRun(parsed);
      break;
    }
Behavior4/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

With no annotations provided, the description carries the full burden of behavioral disclosure. It effectively describes key behaviors: it returns output, token usage, estimated provider cost, and trace URL, and clarifies cost tracking (provider bills, not RelayPlane fees). However, it doesn't mention rate limits, authentication needs, or error handling, leaving some gaps.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is front-loaded with the core purpose, followed by usage context and important behavioral notes. Every sentence adds value: the first states the action, the second provides usage guidance, and the third clarifies cost details. It's concise with zero wasted words.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness4/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Given no annotations and no output schema, the description does well to cover purpose, usage, and key behavioral traits like return values and cost tracking. However, it lacks details on error cases, response format beyond listed items, or performance characteristics, which could be useful for a tool with 4 parameters and no structured output.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters3/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema description coverage is 100%, so the schema already documents all parameters thoroughly. The description adds no additional parameter semantics beyond what's in the schema, such as examples or constraints. Baseline 3 is appropriate since the schema does the heavy lifting.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description clearly states the specific action ('Execute a single AI model call') and resource ('AI model'), distinguishing it from sibling tools like relay_models_list (list models) and relay_workflow_run (execute full workflows). It explicitly mentions testing prompts before building workflows, which helps differentiate its use case.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The description provides explicit guidance on when to use this tool ('Useful for testing prompts before building full workflows'), implying alternatives like relay_workflow_run for production workflows. It also notes cost tracking specifics, helping users understand appropriate contexts for usage versus other tools.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Install Server

Other Tools

Latest Blog Posts

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/RelayPlane/mcp-server'

If you have feedback or need assistance with the MCP directory API, please join our Discord server