# Agent Flow

Inputs, agents, tools, and the flows that connect them. Mermaid marks this type as beta.

## Examples

The playground samples for this type, rendered by Ondine when the documentation builds.

### Example 1

```mermaid
agentflow-beta TB
  brief["Release brief"]@{ shape: input }
  flow writer["Drafting Agent"]
    draft["Draft the notes"]@{ shape: task }
    lookup["changelog_search"]@{ shape: tool }
    guide["Tone of voice"]@{ shape: refdoc }
    draft --> lookup
    draft -.- guide
  end
  flow reviewer["Review Agent"]
    check["Check the claims"]@{ shape: task }
    ok["Accurate?"]@{ shape: decision }
    check --> ok
  end
  publish["Publish"]@{ shape: action }
  brief --> writer
  writer --> reviewer
  ok --> publish

```

### Example 2

```mermaid
---
config:
  theme: default
  look: classic
  layout: dagre
---
agentflow-beta TB
  brief["Release brief"]@{ shape: input }
  flow writer["Drafting Agent"]
    draft["Draft the notes"]@{ shape: task }
    lookup["changelog_search"]@{ shape: tool }
    guide["Tone of voice"]@{ shape: refdoc }
    draft --> lookup
    draft -.- guide
  end
  flow reviewer["Review Agent"]
    check["Check the claims"]@{ shape: task }
    ok["Accurate?"]@{ shape: decision }
    check --> ok
  end
  publish["Publish"]@{ shape: action }
  brief --> writer
  writer --> reviewer
  ok --> publish

```

### Example 3

```mermaid
agentflow-beta LR
  brief["Brief"]@{ shape: input }
  draft["Draft copy"]@{ shape: task }
  spellcheck["spell_check"]@{ shape: tool }
  guide["Style guide"]@{ shape: refdoc }
  publish["Publish"]@{ shape: action }

  brief --> draft --> spellcheck --> publish
  draft -.- guide

```

### Example 4

```mermaid
agentflow-beta TB
  global
    corpus["Shared corpus"]@{ shape: refdoc }
  end

  flow summarizer["Summarizer"]
    summarize["Summarize"]@{ shape: task }
    summarize -.- corpus
  end

  flow indexer["Indexer"]
    index["Build index"]@{ shape: task }
    index -.- corpus
  end

```

### Example 5

```mermaid
agentflow-beta TB
  flow intake["Intake"]
    receive["Receive request"]@{ shape: input }
  end

  flow processing["Processing"]
    validate["Validate"]@{ shape: task }
    enrich["Enrich"]@{ shape: task }
    validate --> enrich
  end
  processing@{ view: "collapsed" }

  flow output["Output"]
    publish["Publish"]@{ shape: action }
  end

  receive --> validate
  enrich --> publish

```

### Example 6

```mermaid
agentflow-beta TB
  connector llm["LLM API"]
  llm@{ protocol: "http", endpoint: "https://api.example.com/chat" }

  flow coffee_team["Coffee Team"]
    city["city"]@{ shape: input, value: "Stockholm" }

    flow researcher["Researcher"]
      research["research_location"]@{ shape: tool, params: "city :: String", returns: "Report" }
      write["write_copy"]@{ shape: tool, connectorRef: "llm.chat", returns: "CoffeeCopy" }
      city --> research --> write
    end
    researcher@{ instruction: "Research the city and draft English coffee copy citing sources." }

    flow designer["Designer"]
      render["generate_html"]@{ shape: tool, connectorRef: "llm.chat", returns: "String" }
      brand["Nordic Brand Guide"]@{ shape: refdoc }
      render -.- brand
    end

    researcher --> designer
  end

```