AgentWorks documentation

Workflow Docs

These docs cover workflow design, execution, and workflow-scoped runtime behavior.

Includes

• Canvas and workflow UX
• Running workflows and workflow manifests
• Step configuration and execution modes
• Learning, evaluation, and monitoring
• Workflow-only interaction patterns such as human feedback and todo_task

Files

• self_improvement_and_reporting.md — start here. System overview of the self-improvement + reporting layer: the Pulse (fix) and Auto-improve (improve) loops, notifications, and the Org dashboard — how workflows self-fix toward their goals and report up so a small team can manage 100+ agents and automations
• auto_improvement_framework.md — metric-backed workflow improvement: metrics as evidence, harden/replan decisions, schedules, and audit logs
• browser_automation.md — durable selector and browser-discovery guidance for workflow browser steps
• cost_and_log_measurement.md — token usage, cost files, phase/run aggregation, and log storage
• deterministic_routing.md — route-by-file routing: deterministic switch, route file producers, and run_workflow route_selections
• evaluation_system.md — workflow evaluation runs, eval step execution, and scoring reports
• human_feedback_system.md — human-in-the-loop requests, UI responses, and Slack notification fallback
• iteration_run_folder_architecture.md — run folder layout and iteration/run isolation
• learn_code_flow.md — learn-code and code-execution modes
• learning_architecture.md — global workflow skill, step-level learning metadata, and saved scripts
• persistent_stores_design.md — workflow db/, report views, knowledgebase updates, and structured persistent outputs
• pre_validation_guide.md — deterministic pre-validation schemas and consistency checks
• step_config_format_specification.md — step_config.json schema and step-level runtime configuration
• tiered_llm_allocation.md — tiered model selection and phase LLM allocation
• todo-task-step-type.md — todo-task orchestration step behavior
• tool_filtering_system.md — workflow tool filtering and runtime tool availability
• workflow_builder_commands_and_tools.md — builder slash commands, privileged tools, and audit flows
• workflow_builder_interactive.md — workflow phase chat and interactive builder internals
• workflow_manifest_architecture.md — manifest format, cleaning rules, and workspace state
• workflow_monitoring.md — run history, costs, reports, and monitoring surfaces
• workflow_scheduling.md — scheduled workflow execution, routing, and history
• workflow_shell_working_directory.md — shell working directories and FolderGuard scope during workflow runs