What is involved?
Add applications, AI providers, agents, gateways, tools, databases, and human checkpoints. Describe each component’s purpose and who is responsible for it.
AGENTICERT FLOW
See the system.
Understand the connections.
Every AI interaction has a system behind it. Flow turns the applications, models, tools, data stores, and people in that system into a map you can explore and explain.
Basic & Advanced viewsInteractive data journeysBrowser-local maps
A SHARED PICTURE OF YOUR AI
Make a complex workflow
easier to talk about.
A feature name rarely explains the whole system. Flow gives product, engineering, security, and business teams a common place to see what connects and ask better questions.
How do the parts interact?
Connect components to describe requests, responses, retrieved context, stored records, or actions. Branches make supporting systems visible alongside the main path.
What do we know?
Keep provider, ownership, data-handling, and security details alongside the map. Distinguish example values, manually declared information, and details that remain unknown.
FROM OVERVIEW TO EXPLANATION
Explore one connection.
Understand the wider story.
Start with a workflow, then use the same map to answer different questions. You do not need to redraw the architecture for every audience.
Choose a starting point.
Open a template or create a map. The template library includes AI chat, customer support, and an enterprise assistant, so you can begin with a recognizable workflow and adapt it.
Inspect the people and services.
Select a component or connection to see its description and recorded details. Use the dashboard’s featured map for an overview, or open the designer to add and edit the parts of your system.
Follow the data.
Switch from Infrastructure to Data journey to focus on information moving between components. Use the guided preview to walk through the connected workflow and discuss each step.
Refine and share a copy.
Edit descriptions, organize the layout, and use undo or redo while working. Export a JSON file to keep a backup or give someone a copy they can import into their own browser.
ONE MAP, TWO LEVELS OF DETAIL
Start simple.
Go deeper when you need to.
Basic and Advanced use the same underlying map. Switching views changes how you explore it and preserves the components, connections, and saved information.
A clear overview for everyone.
Simple service icons and connections make the workflow approachable. Inspect an item, follow the data, or use the essential editing fields without navigating every technical detail at once.
Use it for a product walkthrough, a stakeholder conversation, or your first look at a system.
Room for the supporting detail.
Component cards, connection labels, a larger editing canvas, and detailed fields help you document the architecture. Record dependencies, ownership, information handling, and stated safeguards in context.
Use it when designing a workflow, reviewing dependencies, or preparing a more detailed explanation.
USE THE VIEW THAT ANSWERS THE QUESTION
Built for the conversations
around an AI system.
Product & engineering
Explain where a feature begins, which services it depends on, and what changes when a provider or tool is added. Keep the diagram connected to the information people need to understand it.
Security & governance
Locate data destinations, ownership boundaries, and declared controls. Use unknown information as a review question, then ask the responsible team for the supporting record.
Customers & leadership
Walk through the system in Basic view. Show where information travels and where a person remains involved, without requiring the audience to read every implementation detail.
YOUR FIRST FEW MINUTES IN FLOW
Start with one task.
Build an explanation around it.
- Open the AI chat template.
Find Templates in the workspace and create your own copy. Give the map a name that identifies the feature or task it describes.
- Make it your system.
Rename the application and provider. Add the storage, tools, or human checkpoints involved in your workflow, and connect them to the appropriate components.
- Describe the information.
Record what a connection carries and why. Leave unconfirmed details unknown instead of copying an example value as if it were verified.
- Review, preview, and export.
Check the overview in Basic, inspect the details in Advanced, and follow the journey in Preview. Export your map to keep a backup or share a copy for review.
A little help from Codex.
Prompt Steps provides task-specific prompts you can copy into Codex with the workspace open. Use them to explain a map, think through a data journey, or request an edit. These prompts support your conversation with an external assistant; Flow does not run an embedded chatbot.
A FEW USEFUL ANSWERS
Know what your
map represents.
Where are my maps saved?
Maps are saved in local storage in the browser you use. There is no cloud account, cross-device synchronization, or server backup in this workspace. Export JSON to keep a separate copy. Clearing site data can remove saved maps.
Does Flow connect to the services on the map?
Drawing a connection documents a relationship; it does not connect to the real system, send requests to its AI provider, or enforce an approval step. Flow does not automatically discover infrastructure or monitor service health.
What do the status colors mean?
They display the status recorded for a component or connection. Built-in templates clearly mark example values. Manually declared statuses describe what someone has entered; they are not independent operational checks.
Can I work with another person?
You can export a map and share its JSON file for someone else to import. Each browser works with its own copy. The workspace does not provide simultaneous editing or shared organizational accounts.
How does Flow relate to AI Transparency?
Flow explains the system and its connections. AI Transparency organizes the disclosed practices around that system, such as data handling, oversight, and incident response. A map can provide context for those disclosures, but it does not establish that a safeguard is effective or that a system is certified.
Explore AI TransparencyShould I put production credentials in a map?
No. Describe the role of a service and the relevant information categories without adding passwords, API keys, or tokens. Use only information you are authorized to document and share.
MAKE YOUR AI EASIER TO UNDERSTAND
Your next explanation
can start with a map.
Explore the workspace, choose a template, and follow the connections.