Agentic AI Atlasby a5c.ai
OverviewWikiGraphFor AgentsEdgesSearchWorkspace
/
GitHubDocsDiscord
iiRecord
Agentic AI Atlas · Builder and Product Fit
page:agent-generate-universal-agentic-stack-builder-fita5c.ai
Search record views/
Record · tabs

Available views

II.Record viewspp. 1 - 1
overviewarticlejsongraph
II.
Page overview

page:agent-generate-universal-agentic-stack-builder-fit

Reference · live

Builder and Product Fit overview

Inspect the raw attributes, linked wiki pages, and inbound or outbound graph edges for page:agent-generate-universal-agentic-stack-builder-fit.

PageOutgoing · 9Incoming · 1

Attributes

nodeKind
Page
title
Builder and Product Fit
displayName
Builder and Product Fit
slug
agent-generate/universal-agentic-stack/01-builder-fit
articlePath
wiki/agent-generate/universal-agentic-stack/01-builder-fit.md
article
# Builder and Product Fit This page is for the question people usually ask first: "Where does this framework or product fit?" ## The fast answer Most builder products do not own one single layer. They usually concentrate value in one of these zones: - Layers 2-3: model access and routing - Layers 4-5: agent definition and runtime behavior - Layer 6: extension, deployment, or ecosystem surface - Layers 10-11: interaction and presentation around another runtime The job is to identify the center of gravity, then record what is delegated. ## Typical placements | Thing you are evaluating | Usually lives in | Why | |---|---|---| | Model gateway or inference SDK | Layers 2-3 | It serves or speaks to the model, but does not own the agent loop. | | Agent framework or orchestration library | Layers 4-5 | It defines the agent loop and some runtime behavior. | | Coding agent CLI | Layers 4-6, often with 7-11 around it | It usually owns core, runtime, platform, and some user-facing surface. | | Plugin or skill system | Layer 6 | It is part of the platform and extension contract. | | Local shell runner, container runner, CI job runner | Layer 8 | It is where actions execute. | | Approval policy or filesystem/network restriction system | Layer 9 | It constrains execution. | | Web app, IDE panel, TUI, API surface | Layers 10-11 | It defines actions and presentation. | ## Product-shape heuristics | If the product mainly sells... | Start by checking... | |---|---| | better model access | Layers 2-3 | | better agent authoring | Layers 4-5 | | deployment, extension, or operator workflows | Layer 6 | | safe execution | Layers 8-9 | | better human workflow around an agent | Layers 10-11 | ## LangGraph and custom-agent builders | Stack area | Where it fits | Examples | |---|---|---| | Agent-Core | Graph/state-machine authoring and model/tool routing | `StateGraph`, graph nodes/edges/state, `create_agent`, tool dispatch, terminal routing | | Agent-Runtime | Host process, durable state, checkpointing, tools, interrupts, and streaming | checkpointer/store, thread state, human-in-the-loop gate, host tool registry | | Agent-Platform | Deployed graph operation and ecosystem extension surfaces | LangGraph Platform, LangSmith deployment, `RemoteGraph`, installed plugins, installed skills, launch/config surfaces | ## Builder categories that look similar but are not | Category | Main layer weight | Why people confuse it | |---|---|---| | Agent builder | 4-5 | It often demos through a UI, so people over-credit 10-11 | | Hosted agent platform | 5-6, often 10-11 too | It wraps the builder in deployment and operational surfaces | | IDE integration | 10-11, sometimes 7 | It feels central because it is visible, but it may delegate the actual runtime | | Gateway or router | 2-3 | It sits in the middle of traffic, which makes it look more "agentic" than it is | ## Rules of thumb - If the thing mainly decides the next step, it is probably Layer 4. - If it mainly hosts state, tools, approvals, and streaming around that loop, it is probably Layer 5. - If it mainly manages installed extensions, channels, marketplaces, and launch contracts, it is probably Layer 6. - If a product depends on an IDE, shell, or cloud control plane for workspace or execution, those layers still exist even when they are delegated. - If the product is strongest when embedded into someone else's app, it probably does not own Layers 7-11. - If the product feels complete because it can edit files and run commands, inspect Layers 7-9 carefully rather than assuming the entire stack. ## What not to do - Do not force one product into all 11 layers just because it offers a polished experience. - Do not collapse runtime and platform into one bucket when extension systems matter. - Do not classify "custom agent building" as presentation just because the first touchpoint is a visual editor. - Do not confuse hosted operation with model ownership. A hosted platform may still delegate Layers 1-3. ## Next pages - Read [`04-worked-examples.md`](./04-worked-examples.md) for concrete placements. - Read [`05-common-confusions.md`](./05-common-confusions.md) if the terminology still feels slippery.
documents
  • layer:3-transport
  • layer:4-agent-core
  • layer:5-agent-runtime
  • layer:6-agent-platform
  • layer:7-workspace
  • layer:8-execution
  • layer:9-sandbox
  • layer:10-interaction
  • layer:11-presentation

Outgoing edges

documents9
  • layer:3-transport·LayerTransport
  • layer:4-agent-core·LayerAgent-Core
  • layer:5-agent-runtime·LayerAgent-Runtime
  • layer:6-agent-platform·LayerAgent-Platform
  • layer:7-workspace·LayerWorkspace
  • layer:8-execution·LayerExecution
  • layer:9-sandbox·LayerSandbox
  • layer:10-interaction·LayerInteraction
  • layer:11-presentation·LayerPresentation

Incoming edges

contains_page1
  • page:agent-generate-universal-agentic-stack·PageUniversal Agentic Stack

Related pages

No related wiki pages for this record.

Shortcuts

Open in graph
Browse node kind