Skip to main content
Glama

run_workflow_step

Execute the next step in an active workflow session by following provided instructions precisely to advance structured multi-step processes.

Instructions

Execute the next step in an active workflow session.

IMPORTANT: When executing workflow steps, follow these rules:

  1. DO NOT provide commentary between workflow steps unless explicitly requested

  2. Simply execute each step according to the workflow instructions

  3. Move immediately to the next step after completing the current one

  4. Only provide output when the workflow specifically requires it (e.g., notify actions, final results)

  5. Focus solely on executing the workflow actions as defined

The tool will provide step-by-step instructions that should be followed exactly.

Input Schema

TableJSON Schema
NameRequiredDescriptionDefault
execution_idYes
step_resultNo
next_step_neededYes

Implementation Reference

  • The core handler function for the 'run_workflow_step' tool. It processes the input arguments, manages the workflow session state, stores step results, handles branching logic, determines the next step, generates execution instructions, or completes the workflow if finished.
    private async runWorkflowStep(args: unknown) {
      const parsed = RunWorkflowStepSchema.parse(args);
      
      const session = this.sessions.get(parsed.execution_id);
      if (!session) {
        throw new Error(`No active workflow session found: ${parsed.execution_id}`);
      }
    
      const workflow = await this.storage.get(session.workflow_id);
      if (!workflow) {
        throw new Error(`Workflow not found: ${session.workflow_id}`);
      }
    
      // Store result from previous step if provided
      if (parsed.step_result !== undefined && session.current_step_index > 0) {
        const previousStep = workflow.steps[session.current_step_index - 1];
        if (previousStep.save_result_as) {
          // Store in step outputs for dependency tracking
          session.step_outputs[previousStep.id] = {
            variable_name: previousStep.save_result_as,
            value: parsed.step_result
          };
          
          // Update previous_variables to track what was before this change
          session.previous_variables = { ...session.variables };
          
          // Update current variables
          session.variables[previousStep.save_result_as] = parsed.step_result;
        }
      }
    
      // Check if workflow is complete
      if (!parsed.next_step_needed || session.current_step_index >= workflow.steps.length) {
        session.status = 'completed';
        this.sessions.delete(parsed.execution_id);
        
        return {
          content: [
            {
              type: 'text',
              text: JSON.stringify({
                status: 'completed',
                execution_id: parsed.execution_id,
                message: `Workflow "${workflow.name}" completed successfully`,
                final_variables: session.variables,
              }, null, 2),
            },
          ],
        };
      }
    
      // Check if the current step was a branch step
      const currentStep = workflow.steps[session.current_step_index];
      let nextStepIndex = session.current_step_index + 1;
      
      // Handle branching logic
      if (currentStep.action === 'branch' && parsed.step_result) {
        // Parse the branch result to find the target step
        const branchResult = parsed.step_result.toString();
        const gotoMatch = branchResult.match(/step (\d+)/i);
        if (gotoMatch) {
          const targetStep = parseInt(gotoMatch[1]);
          // Find the index of the target step (steps are 1-indexed, array is 0-indexed)
          nextStepIndex = workflow.steps.findIndex(s => s.id === targetStep);
          if (nextStepIndex === -1) {
            throw new Error(`Branch target step ${targetStep} not found`);
          }
        }
      }
      
      // Update session with next step
      session.current_step_index = nextStepIndex;
      
      if (session.current_step_index >= workflow.steps.length) {
        session.status = 'completed';
        this.sessions.delete(parsed.execution_id);
        
        return {
          content: [
            {
              type: 'text',
              text: JSON.stringify({
                status: 'completed',
                execution_id: parsed.execution_id,
                message: `Workflow "${workflow.name}" completed successfully`,
                final_variables: session.variables,
              }, null, 2),
            },
          ],
        };
      }
    
      // Generate instructions for the next step
      const nextStep = workflow.steps[session.current_step_index];
      const stepInstructions = this.generateStepInstructions(workflow, nextStep, session);
    
      return {
        content: [
          {
            type: 'text',
            text: stepInstructions,
          },
        ],
      };
    }
  • Zod schema validating the input for run_workflow_step: execution_id (string), step_result (any, optional), next_step_needed (boolean). Used in handler and tool registration.
    const RunWorkflowStepSchema = z.object({
      execution_id: z.string(),
      step_result: z.any().optional(),
      next_step_needed: z.boolean(),
    });
  • src/index.ts:282-295 (registration)
    Tool registration in the getTools() method, defining name 'run_workflow_step', its description, and input schema converted to JSON schema.
          {
            name: 'run_workflow_step',
            description: `Execute the next step in an active workflow session.
    
    IMPORTANT: When executing workflow steps, follow these rules:
    1. DO NOT provide commentary between workflow steps unless explicitly requested
    2. Simply execute each step according to the workflow instructions
    3. Move immediately to the next step after completing the current one
    4. Only provide output when the workflow specifically requires it (e.g., notify actions, final results)
    5. Focus solely on executing the workflow actions as defined
    
    The tool will provide step-by-step instructions that should be followed exactly.`,
            inputSchema: zodToJsonSchema(RunWorkflowStepSchema),
          },
  • src/index.ts:134-135 (registration)
    Dispatcher case in the central CallToolRequestSchema handler that routes 'run_workflow_step' calls to the runWorkflowStep method.
    case 'run_workflow_step':
      return await this.runWorkflowStep(args);
  • Key helper function called by the handler to generate detailed step-by-step execution instructions, including template variable resolution, dependency-based visible variables, change highlighting, and upcoming steps preview.
    private generateStepInstructions(workflow: Workflow, step: Step, session: WorkflowSession): string {
      const lines: string[] = [];
      
      lines.push('=== WORKFLOW STEP EXECUTION ===');
      lines.push('');
      lines.push(`Workflow: ${workflow.name} (${session.execution_id})`);
      lines.push(`Step ${session.current_step_index + 1} of ${session.total_steps}`);
      lines.push('');
      lines.push(`Action: ${step.action.toUpperCase()}`);
      lines.push(`Description: ${this.resolveTemplateVariables(step.description, session.variables)}`);
      lines.push('');
      
      // Step-specific instructions
      if (step.action === 'tool_call' && 'tool_name' in step && 'parameters' in step) {
        lines.push('Execute the following tool:');
        lines.push(`Tool: ${step.tool_name}`);
        // Resolve template variables in parameters
        const resolvedParams = this.resolveTemplateInObject(step.parameters, session.variables);
        lines.push(`Parameters: ${JSON.stringify(resolvedParams, null, 2)}`);
      } else if (step.action === 'analyze' || step.action === 'consider' || step.action === 'research' || 
                 step.action === 'validate' || step.action === 'summarize' || step.action === 'decide' ||
                 step.action === 'extract' || step.action === 'compose') {
        lines.push(`Perform the cognitive action: ${step.action}`);
        if ('input_from' in step && step.input_from) {
          lines.push('Using input from:');
          for (const varName of step.input_from) {
            if (session.variables[varName] !== undefined) {
              lines.push(`  ${varName}: ${JSON.stringify(session.variables[varName])}`);
            }
          }
        }
        if ('criteria' in step && step.criteria) {
          lines.push(`Criteria: ${this.resolveTemplateVariables(step.criteria, session.variables)}`);
        }
      } else if (step.action === 'wait_for_input' && 'prompt' in step) {
        lines.push('Request input from the user:');
        lines.push(`Prompt: ${this.resolveTemplateVariables(step.prompt, session.variables)}`);
        if ('input_type' in step) {
          lines.push(`Expected type: ${step.input_type}`);
        }
      } else if (step.action === 'transform' && 'transformation' in step) {
        lines.push('Apply transformation:');
        lines.push(this.resolveTemplateVariables(step.transformation, session.variables));
        if ('input_from' in step && step.input_from) {
          lines.push('To variables:');
          for (const varName of step.input_from) {
            if (session.variables[varName] !== undefined) {
              lines.push(`  ${varName}: ${JSON.stringify(session.variables[varName])}`);
            }
          }
        }
      } else if (step.action === 'branch' && 'conditions' in step) {
        lines.push('Evaluate conditions and determine next step:');
        lines.push('');
        for (const condition of step.conditions) {
          lines.push(`IF ${condition.if} THEN GOTO step ${condition.goto_step}`);
        }
        lines.push('');
        lines.push('Evaluate the conditions using the current variables and respond with:');
        lines.push('- Which condition is true');
        lines.push('- The step number to jump to (e.g., "Branching to step 8")');
      } else if (step.action === 'notify' && 'message' in step) {
        lines.push('Send notification:');
        lines.push(`Message: ${this.resolveTemplateVariables(step.message, session.variables)}`);
        if ('channel' in step && step.channel) {
          lines.push(`Channel: ${this.resolveTemplateVariables(step.channel, session.variables)}`);
        }
      }
      
      if (step.save_result_as) {
        lines.push('');
        lines.push(`Save the result as: ${step.save_result_as}`);
      }
      
      // Get visible variables based on dependencies
      const visibleVariables = this.getVisibleVariables(workflow, step, session);
      
      // Show variable changes if we have previous state
      if (session.current_step_index > 0 && Object.keys(session.previous_variables).length > 0) {
        const changes = this.getChangedVariables(session);
        
        if (changes.added.length > 0 || changes.modified.length > 0) {
          lines.push('');
          lines.push('Variable changes:');
          
          for (const varName of changes.added) {
            if (varName in visibleVariables) {
              lines.push(`  + ${varName}: ${JSON.stringify(session.variables[varName])}`);
            }
          }
          
          for (const varName of changes.modified) {
            if (varName in visibleVariables) {
              lines.push(`  ~ ${varName}: ${JSON.stringify(session.variables[varName])}`);
            }
          }
        }
      }
      
      // Show current visible variables
      lines.push('');
      lines.push('Available variables:');
      const visibleVarNames = Object.keys(visibleVariables);
      if (visibleVarNames.length === 0) {
        lines.push('  (none - use dependencies to access previous step outputs)');
      } else {
        for (const [key, value] of Object.entries(visibleVariables)) {
          lines.push(`  ${key}: ${JSON.stringify(value)}`);
        }
      }
      
      // Show upcoming steps (progressive loading)
      if (session.current_step_index < workflow.steps.length - 1) {
        lines.push('');
        lines.push('Upcoming steps:');
        const maxPreview = Math.min(session.current_step_index + 3, workflow.steps.length);
        for (let i = session.current_step_index + 1; i < maxPreview; i++) {
          const upcomingStep = workflow.steps[i];
          lines.push(`  ${i + 1}. ${upcomingStep.action}: ${upcomingStep.description}`);
        }
        if (maxPreview < workflow.steps.length) {
          lines.push(`  ... and ${workflow.steps.length - maxPreview} more steps`);
        }
      }
      
      lines.push('');
      lines.push('After completing this step, call run_workflow_step with:');
      lines.push('- execution_id: ' + session.execution_id);
      lines.push('- step_result: <the result of this step, if any>');
      lines.push('- next_step_needed: true (or false if the workflow should end)');
      
      return lines.join('\n');
    }
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 behavioral traits: it's an execution tool (implying mutation), requires following specific rules (e.g., no commentary, immediate progression), and outlines output behavior (only when required). It doesn't cover aspects like error handling or permissions, but provides substantial context.

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

Conciseness4/5

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

The description is well-structured with a clear purpose statement followed by bullet-point rules, making it front-loaded and easy to parse. It's appropriately sized, though the rule list could be slightly condensed without losing clarity.

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

Completeness3/5

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

Given the complexity of workflow execution, no annotations, and no output schema, the description provides good behavioral rules but lacks parameter explanations and details on return values or error cases. It's partially complete but has notable gaps for a tool with 3 parameters and mutation implications.

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

Parameters2/5

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

The schema description coverage is 0%, and the description provides no information about the 3 parameters (execution_id, step_result, next_step_needed). It mentions 'step-by-step instructions' but doesn't explain how parameters relate to execution, leaving semantics unclear and failing to compensate for the schema gap.

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

Purpose4/5

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

The description clearly states the verb 'execute' and resource 'next step in an active workflow session', making the purpose understandable. However, it doesn't explicitly differentiate from sibling tools like 'start_workflow' or 'rollback_workflow', which might handle workflow execution in different contexts.

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 usage rules (e.g., 'DO NOT provide commentary between workflow steps', 'Move immediately to the next step after completing the current one'), specifying when and how to use this tool versus alternatives. This offers clear operational guidance beyond basic functionality.

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/FiveOhhWon/workflows-mcp'

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