🇺🇸English

Agently LangGraph To TriggerFlow

This skill covers migration from LangGraph orchestration concepts into Agently TriggerFlow. It focuses on StateGraph, graph state, interrupts, persistence, streaming, subgraphs, and durable execution. It does not cover high-level LangChain agent migration unless that agent layer is secondary to the orchestration problem.

Prerequisite: Agently >= 4.0.8.5.

Scope

Use this skill for:

• StateGraph
• nodes and edges
• Send or dynamic worker fan-out
• Command-style resume or control flow
• graph state
• interrupts
• persistence or checkpoints
• streaming
• subgraphs
• durable execution
• LangGraph-based human-in-the-loop behavior

Do not use this skill for:

• one-request agent migration as the main problem
• direct model-provider setup as the main problem
• non-migration TriggerFlow implementation work that already starts from Agently

Workflow

1. Start with references/concept-map.md to translate LangGraph orchestration concepts into the current TriggerFlow skill tree.
2. Read references/common-migration-recipes.md when the source graph resembles a standard LangGraph production shape.
3. Read references/non-1to1-differences.md when the migration should change design rather than mimic syntax.
4. Read references/current-skill-map.md to choose the target TriggerFlow skill combination.
5. If the source problem is actually high-level LangChain agent behavior, switch to agently-langchain-to-agently.
6. If behavior still looks wrong, use references/troubleshooting.md.

Core Mental Model

LangGraph and TriggerFlow are both orchestration layers, but they are not shaped the same way.

LangGraph emphasizes:

• graph state
• nodes and edges
• interrupts plus checkpoints
• thread and checkpoint persistence

TriggerFlow emphasizes:

• signals and handlers
• runtime data and flow data
• explicit runtime stream
• config plus execution-state restore

So the migration target is usually:

• concept translation
• state-boundary redesign
• persistence-boundary redesign

not a syntax-by-syntax rewrite.

Selection Rules

• StateGraph, core node and edge wiring -> agently-triggerflow-orchestration
• routers, Send-like fan-out, loops, evaluator-reviser, or handoff-like workflow shapes -> agently-triggerflow-patterns
• graph state, shared state, or dependency boundaries -> agently-triggerflow-state-and-resources
• subgraphs or isolated child workflow boundaries -> agently-triggerflow-subflows
• streaming model output inside the graph -> agently-triggerflow-model-integration
• definition export/import or graph visualization -> agently-triggerflow-config
• checkpoint-like restore of a running instance -> agently-triggerflow-execution-state

Important Boundaries

• LangGraph threads and checkpointers do not map literally to one TriggerFlow API
• LangGraph Send and Command do not map literally to one TriggerFlow primitive; keep the worker pattern or resume intent, then redesign the flow
• durable restore in Agently usually becomes config + execution-state, and sometimes also session-memo
• graph state does not map literally to one TriggerFlow state container; choose between runtime_data, flow_data, resources, and external storage
• streaming modes do not map literally one-to-one; route model streaming and runtime stream separately

References

• references/source-map.md
• references/concept-map.md
• references/common-migration-recipes.md
• references/non-1to1-differences.md
• references/current-skill-map.md
• references/troubleshooting.md

Weekly Installs

1

Repository

agentera/agently-skills

First Seen

1 day ago

Security Audits

Gen Agent Trust HubPassSocketPassSnykPass

Installed on

mcpjam1

claude-code1

replit1

junie1

windsurf1

zencoder1