RelayPlane

Instructions

Implementation Reference

• src/tools/relay-run-get.ts:44-77 (handler)

  Core implementation of the 'relay_run_get' tool handler. Fetches run data by ID, formats timestamps and trace URL, handles missing runs.

  export async function relayRunGet(
    input: RelayRunGetInput
  ): Promise<RelayRunGetResponse> {
    const run = getRunById(input.runId);
    const config = getConfig();
  
    if (!run) {
      return {
        found: false,
        error: `Run with ID "${input.runId}" not found. Note: Run history is stored in memory and clears on server restart.`,
      };
    }
  
    return {
      found: true,
      run: {
        runId: run.runId,
        type: run.type,
        name: run.type === 'workflow' ? run.workflowName : undefined,
        model: run.type === 'single' ? run.model : undefined,
        success: run.success,
        startTime: run.startTime.toISOString(),
        endTime: run.endTime.toISOString(),
        durationMs: run.durationMs,
        usage: run.usage,
        input: run.input,
        output: run.output,
        steps: run.steps,
        error: run.error,
        traceUrl: `${config.traceUrlBase}/${run.runId}`,
        contextReduction: run.contextReduction,
      },
    };
  }

• src/tools/relay-run-get.ts:11-42 (schema)

  Input schema (Zod), input type, and response interface for the tool.

  export const relayRunGetSchema = z.object({
    runId: z.string().describe('The run ID to retrieve'),
  });
  
  export type RelayRunGetInput = z.infer<typeof relayRunGetSchema>;
  
  export interface RelayRunGetResponse {
    found: boolean;
    run?: {
      runId: string;
      type: 'single' | 'workflow';
      name?: string;
      model?: string;
      success: boolean;
      startTime: string;
      endTime: string;
      durationMs: number;
      usage: {
        promptTokens: number;
        completionTokens: number;
        totalTokens: number;
        estimatedProviderCostUsd: number;
      };
      input?: any;
      output?: any;
      steps?: Record<string, any>;
      error?: string;
      traceUrl: string;
      contextReduction?: string;
    };
    error?: string;
  }

• src/tools/relay-run-get.ts:79-92 (schema)

  MCP tool definition including name, description, and JSON schema for input validation.

  export const relayRunGetDefinition = {
    name: 'relay_run_get',
    description: 'Get full details of a specific run including all step outputs and trace URL.',
    inputSchema: {
      type: 'object' as const,
      properties: {
        runId: {
          type: 'string',
          description: 'The run ID to retrieve',
        },
      },
      required: ['runId'],
    },
  };

• src/server.ts:59-67 (registration)

  Registers the tool by including its definition in the TOOLS array returned by listTools MCP handler.

  const TOOLS = [
    relayModelsListDefinition,
    relayRunDefinition,
    relayWorkflowRunDefinition,
    relayWorkflowValidateDefinition,
    relaySkillsListDefinition,
    relayRunsListDefinition,
    relayRunGetDefinition,
  ];

• src/server.ts:139-142 (handler)

  Server-side dispatch for the tool call: parses input with schema and invokes the handler function.

  case 'relay_run_get': {
    const parsed = relayRunGetSchema.parse(args);
    result = await relayRunGet(parsed);
    break;