Volley / Tools / AI assistants
AI sequence diagram generatorNo account
Let your AI draw the sequence diagram.
Describe a flow to ChatGPT, Claude or Gemini, or point a coding agent at your code. You get back a Volley diagram with the request and response on each arrow, and a link that opens it in the editor.
In any chat assistant
Paste this prompt into ChatGPT, Claude, Gemini or any assistant that can read a web page, and fill in the flow at the end. The assistant reads the syntax primer and replies with a link that opens the diagram in the editor. It works without an account.
Draw this flow as a Volley sequence diagram.
Read https://volley.sh/llms.txt for the syntax.
Put each call's request and response in in: and out: blocks,
and write <redacted> for secrets. Reply with the source in a
code block and a link: https://volley.sh/edit#text= followed by
the percent-encoded source.
The flow: [describe it, or paste an API doc, a log or the code]If your assistant cannot open web pages, paste the contents of /llms.txt into the chat instead.
In a coding agent
Connect Claude Code, Cursor or another MCP client to Volley. The agent writes the diagram from your code, checks it for errors and fixes them, then saves it and gives you a short link.
claude mcp add --transport http volley https://volley.sh/mcp \
--header "Authorization: Bearer $VOLLEY_KEY"Then ask for the flow:
Read the checkout code in src/payments and draw the flow from the
Buy button to fulfilment as a Volley diagram, with the real request
and response bodies. Check it with check_diagram, fix any errors,
then save it and give me the link.Create the key under Connect an agent in the sidebar after you sign in. The MCP docs list every tool and show the JSON config for other clients.
What the AI writes
This is the source the assistant writes. It is plain text, so you can read it, fix it by hand and review changes to it in a pull request:
title: Exchange a code for a token
App -> Auth Server: POST /token
in: Token request
grant_type: authorization_code
code: SplxlOBeZQQYbYS6WxSbIA
code_verifier: dBjftJeZ4CVP-mB92K27uhbUJU1p1r_wW1gFWFOEjXk
Auth Server --> App: 200 OK
out: Token response
```json
{ "access_token": "eyJhbGciOi...", "token_type": "Bearer", "expires_in": 3600 }
```Good prompts to try
- "Draw the OAuth flow this app uses to sign in, from the login button to the first API call."
- "Turn this webhook handler into a diagram of every event it handles and what it calls next."
- "Here is a failing request log. Draw it and mark the call that failed."
- "Diagram how a request moves through these three services, with the retry when the payment API times out."
Or start from a template and ask the assistant to change it to match your integration.
Questions
Which AI assistants can write Volley diagrams?
Chat assistants such as ChatGPT, Claude and Gemini can learn the syntax from /llms.txt and give you a link that opens in the editor. Coding agents such as Claude Code and Cursor connect to the remote MCP server, which lets them check and save diagrams themselves.
Do I need an account?
Not for diagrams a chat assistant writes: the link carries the whole diagram, and the editor is free with no sign-in. The MCP server and HTTP API need an API key, which comes with a free account and includes 200 calls a month. Pro includes 20,000.
Is the diagram uploaded when I open the link?
No. The diagram travels in the part of the URL after the #, which browsers never send to a server. It is stored only if you choose to save it.
What if the assistant gets the syntax wrong?
The editor marks each error with its line and column, and so do the MCP server, HTTP API and CLI. An agent with the MCP server checks its own diagram and fixes it before you see it. In a chat, paste the error back and ask for a fix.
Why a sequence diagram with payloads instead of a plain one?
An agent reading your code can see what each call sends, so it can write that down too. In Volley the request and response sit on the arrow, collapsed until you click it, so you can check what step 4 sends without the diagram turning into a wall of JSON.