Custom graph behind an app tool
A LangGraph workflow can produce an analysis while General Augment governs how the product assistant invokes it.
Category comparison
LangGraph and LangSmith vs an app-agent backend
Use LangGraph and LangSmith when your team wants to build, operate, trace, and evaluate custom agent workflows. Use General Augment when the priority is embedding a governed agent into an existing app with user identity, memory, tools, approvals, channels, usage, and support evidence.
Use LangGraph and LangSmith when
- Your team needs graph-based orchestration, durable execution, streaming, human-in-the-loop, and persistence.
- You want tracing, evaluation, prompt, and deployment workflows across custom agents.
- The agent system is a core engineering project your team will operate.
Use General Augment when
- The agent should be a product capability, not a new platform project.
- You need app-user identity, memory, governed tools, approvals, channels, usage, and traces in one self-serve backend.
- You want to keep custom orchestration services as optional specialist backends.
Decision table
Compare the layer, not just the feature list.
Decision area
LangGraph and LangSmith
General Augment
Runtime
LangGraph focuses on durable orchestration, streaming, human-in-the-loop, and persistence.
General Augment focuses on the product backend around agent turns and app integration.
Observability
LangSmith is strong for tracing, evaluation, prompts, and agent debugging across frameworks.
General Augment gives app operators run evidence tied to users, tools, approvals, channels, and usage.
Buyer
Best for engineering teams building custom agent systems.
Best for product teams adding managed agents to existing apps.
Architecture fit
Where the categories fit together.
LangGraph and LangSmith can be engineering infrastructure for custom agents. General Augment is the app-agent backend that packages agent behavior for existing products.
Layered architecture
Product UI
-> app backend
-> General Augment project
-> governed tool
-> LangGraph workflow
-> trace, approval, usage, and responseSources checked
Migration path
Start where you are. Add the agent backend when the product needs it.
- 01Use LangGraph when the workflow needs explicit graph control or durable custom orchestration.
- 02Use LangSmith when traces and evals are the center of the engineering workflow.
- 03Put customer-facing product-agent entrypoints behind General Augment when app users, channels, memory, approvals, and usage need to be packaged.
- 04Connect LangGraph services back to General Augment through governed tools when useful.
Specific examples
Product support agent
General Augment keeps the app-user context, memory, approvals, and traces available to product and support teams.
Evaluation workflow
LangSmith can remain part of an engineering improvement loop, while General Augment owns the customer-facing backend surface.
When not to use General Augment
Use the smaller tool when the smaller tool is enough.
Use LangGraph and LangSmith directly when custom graph orchestration is your core differentiator.
Use LangSmith directly when your primary problem is tracing and evaluation across a custom agent fleet.
Use General Augment later if product packaging, channels, app users, and governed tool surfaces become the bottleneck.
FAQ
Does General Augment replace LangGraph?
No. LangGraph is useful for custom orchestration. General Augment is useful when app teams need a governed backend around product-agent turns.
Does General Augment replace LangSmith?
No. LangSmith is strong for tracing and evaluation workflows. General Augment focuses on app-user memory, tools, approvals, channels, usage, and operational evidence for product agents.
Can LangGraph workflows run behind General Augment?
Yes. A LangGraph workflow can be exposed as a governed tool or connector action when the product assistant should use it.