This MCP server provides comprehensive documentation for the Stabilize workflow engine, dynamically served from the installed library (stabilize v0.19.1).
Streamable HTTP transport (recommended):
URL: https://mcp.stabilize.rodmena.ai/mcp
SSE transport (legacy):
URL: https://mcp.stabilize.rodmena.ai/sse
Getting started guide: critical rules (disk-based SQLite, built-in tasks) and a complete working example to copy as a starting template
Shell task documentation: context parameters, outputs, upstream substitution, secrets
HTTP task documentation: methods, auth, uploads, downloads, retries
Python task documentation: script/module/function modes, INPUT/RESULT convention
Docker task documentation: run/exec/build/pull actions, resource limits, GPU
SSH task documentation: remote command execution, key auth, timeouts
Long-running tasks: servers, auto-restart, restart_on_failure vs continue_on_failure
Core classes API: Workflow, StageExecution, TaskExecution, WorkflowStatus
Task implementation: Task interface, TaskResult factory methods, RetryableTask, SkippableTask, jump_to, builder pattern
Task registry: registering tasks, aliases, implementing_class matching
DAG patterns: sequential, parallel fan-out/fan-in, complex multi-level DAGs
Context and outputs data flow between stages, upstream output merging
Common mistakes and how to fix them (12 pitfalls)
Complete example: sequential pipeline with validation, processing, error handling
Complete example: parallel stages with fan-out and join aggregation
Complete imports reference: core, advanced, verification, conditions, assertions, validation, errors, agentic toolkit
Verification system: VerifyResult, OutputVerifier, custom verifiers for output validation
Structured conditions (Kubernetes-style): Condition, ConditionSet, status tracking
Assertion helpers: context/output/config assertions, exception hierarchy
Configuration validation: JSON Schema-based context validation, built-in schemas
Error handling & reliability: transient/permanent errors, retries, deduplication, crash recovery, stateful retries
Synthetic stages: before/after/onFailure stages, StageDefinitionBuilder, execution order
Concurrency & race conditions: optimistic locking, StartStage race handling, best practices
Event sourcing: setup, subscriptions, replay, projections, event types, event stores
Advanced workflow control-flow patterns (20 of 43 WCPs): joins, splits, signals, mutex, milestones, loops, sub-workflows
Agentic workflows: LLMClient model client, one-shot LLMTask, tool-calling AgentLoopTask (ReAct), @tool/ToolRegistry, durable human-in-the-loop approvals (ApprovalTask + approve/reject), live streaming (WorkflowStream + emit_progress), fan-in output_reducers, quorum/discriminator patterns, complete runnable multi-agent template
Search all Stabilize documentation for a specific topic or keyword
The complete Stabilize reference - everything `stabilize prompt` outputs (~25k tokens). Prefer the targeted section tools unless you need the whole reference at once
A product of Rodmena Ltd