You can connect an external AI assistant like Claude and ChatGPT to Ava and use it to read your Spaces, product gaps and open deals, or to consult Ava about one of your Spaces. Ava answers using what it knows about that Space: the accounts, the calls, and what it has learned over time. You get Ava's answer without leaving the assistant you were already working in.
Availability
Available to any Ava user with a Standard or Lite license.
You connect with your own Ava sign-in, so you see only what you can already see in Ava.
Server address: https://mcp.ava.vivun.com/mcp
Supported today: Claude (desktop app and Claude Code), ChatGPT, Slackbot, and Microsoft Copilot.
Ava is not listed in the Claude or ChatGPT connector directory yet, so you add it by its server address, as described below.
Before you start
You need an Ava account you can already sign in to, and at least one Space. If you have no Spaces yet, there is nothing for Ava to answer from.
If your company uses Claude Team or Enterprise, an organization Owner may need to add the Ava connector for your organization before you can connect (see For Claude Team and Enterprise owners below).
Claude
Connect Claude to Ava
In the Claude desktop app, open Settings > Connectors.
If Ava already appears in the list (because your organization added it), click Connect and skip to step 5.
Otherwise, click Add custom connector.
Enter a name such as Ava, paste https://mcp.ava.vivun.com/mcp as the URL, and click Add. Leave the advanced settings empty.
Click Connect. A browser window opens on Ava's sign-in page. Sign in the same way you sign in to the Ava web application.
Review the access request and click Allow. See About the sign-in warning below.
Return to Claude. If Ava does not appear right away, quit and reopen Claude.
In a new chat, make sure Ava is turned on in the tools menu, then ask something like "Ask Ava what the latest is on the Acme Space."
Connect Claude Code to Ava
In a terminal, run:
claude mcp add --transport http ava-mcp https://mcp.ava.vivun.com/mcpStart Claude Code, run
/mcp, select ava-mcp, and choose Authenticate.Sign in to Ava in the browser window that opens and click Allow.
For Claude Team and Enterprise owners
On Claude Team and Enterprise plans, an Owner adds the connector once for the whole organization:
In Claude, open Admin settings > Connectors.
Click Add custom connector, name it Ava, enter https://mcp.ava.vivun.com/mcp, and save.
Each member then connects and signs in with their own Ava account from Settings > Connectors. Adding the connector does not give anyone access to Ava data they could not already see.
ChatGPT
Connect ChatGPT to Ava
A ChatGPT workspace admin adds Ava once for the whole workspace, and members can then use it in any ChatGPT conversation. This is available on ChatGPT Enterprise, Team and Edu workspaces. Setup takes about five minutes and does not need anything from Vivun.
Before you start: you must be a ChatGPT workspace admin, and your organization needs an active Ava subscription. Each person who will use Ava in ChatGPT needs their own Ava account.
For ChatGPT workspace admins
In ChatGPT, open the Admin Console and go to Plugins.
Click Add, then Create New.
Enter the following:
Name:
AvaDescription:
Ask Ava about your accounts, deals and spaces.Connection:
https://mcp.ava.vivun.com/mcpAuthentication: OAuth. Leave client registration on automatic and do not enter a Client ID or Client secret.
Create the plugin. ChatGPT generates the OAuth values itself. Check that the Client ID is a
https://chatgpt.com/oauth/.../client.jsonaddress and the Callback URL is ahttps://chatgpt.com/connector/oauth/...address. If either looks different, the Connection address was entered incorrectly. Re-enter it exactly as shown above.Turn the plugin on for the whole workspace, or only for the groups that should use Ava.
For ChatGPT users
In a ChatGPT conversation, choose Ava from your available plugins.
The first time, a sign-in window opens. Sign in the same way you sign in to the Ava web application, and approve the access request. You stay signed in afterwards.
Ask something like "Ask Ava what the latest is on the Acme Space."
If you sign in with Salesforce, see About the sign-in warning below.
What ChatGPT can see
Ava only returns what the person asking can already see in Ava. The plugin gives no one new access.
Questions asked from ChatGPT do not appear as Ava chats and do not change what Ava remembers about your accounts, as described in Consultations are kept apart from your work in Ava below.
To stop everyone using Ava in ChatGPT, turn the plugin off in the Admin Console. To remove one person, remove their Ava account.
Slack
Connect Slackbot to Ava
Slackbot needs a one-time setup by a Slack administrator. See Connect Ava to Slackbot.
Microsoft Copilot
Connect Microsoft Copilot Studio to Ava
A Copilot Studio maker adds Ava as a tool in a Copilot Studio agent. The setup needs a Client ID and Client secret from Vivun, and Vivun registers your environment's Redirect URL partway through. See Connect Ava to Microsoft Copilot Studio.
About the sign-in warning
If you sign in to Ava with Salesforce, Salesforce shows an Allow Access? screen with a Security Warning the first time you connect. Salesforce shows this warning whenever an app asks to act on your behalf. It is expected when you set up Ava yourself.
The request comes from Vivun Ava Prod Auth0, Ava's sign-in service.
Only click Allow if you started the connection yourself from your AI assistant. As the warning says, do not proceed if someone contacted you and told you to.
Allowing it does not give your assistant access to anything you cannot already see in Ava, and it cannot change your CRM (see below).
What you can ask for
Your list of Spaces. Your assistant can list the Spaces you have access to. Each Space shows its name, the Salesforce account ID and account owner, and the same last-activity date and reason you see on the Spaces screen in Ava. The list can be filtered by owner or by recent activity. Each Space also says whether its data is still loading or ready, with the number of calls linked and summarized and the date of the latest call, so you can tell when a new Space is ready to ask about.
An answer about one Space. Ask a question about a Space, and Ava works through that Space's accounts, calls and memory to answer it. Most answers arrive within 15 to 40 seconds.
Product gaps for your Spaces. Your assistant can read the product gaps Ava has recorded for up to 50 Spaces in one request. Each gap lists where it came from, such as a call, a Slack message or an uploaded file, with the date. A Space you can't read is marked as not available, and the other Spaces still come back.
Your organization's Training knowledge (Trainers and Admins). Your assistant can browse what Ava has learned in Training, filter it by product line, type or date, and read the sources behind an entry. Each entry shows the entries it is linked to, with each linked entry listed once. Trainers and Admins can also teach Ava from their assistant. See Teach Ava from your assistant below.
Your open deals. Your assistant can read a page of your open deals from Salesforce, ordered by close date. Each deal shows the account, stage and owner, plus the columns you have set up in Ava. By default you see deals for everyone on your Spaces. Ask for "my deals" to see only the ones you own. Deals that have slipped past their close date are included.
Search your Salesforce accounts. Your assistant can search your Salesforce accounts, using the same search as the Add Space window in Ava. Each result shows whether a Space already exists for the account and whether it is already in your list.
Add a Space. When you ask, or confirm, your assistant can add an account's Space for you. If the Space is already in your list, nothing changes. If the account already has a Space in your organization, adding it joins you to that Space instead of creating a new one.
Two ready-made reports. They appear in your assistant's prompt menu (a slash command in Claude), so you get the same structure every time. In Claude's chat (web or desktop), start a new chat and select + in the message box. Hover Connectors, scroll down inside the list to the "Add from…" rows, and hover Add from Ava MCP. Choose Forecast report or Product gaps summary, fill in the optional fields or leave them blank to cover everything, select Add prompt, then send.
Forecast report. You can limit it to your own deals or to a range of close dates. Every number comes from your live Salesforce deal data, not from the assistant's own wording. It totals the deals by forecast category, with a count for each. It treats a deal as at risk when its close date has passed or it has no recorded activity, then asks Ava about the 10 largest at-risk deals, using the last 30 days of calls and messages. It ends with a table of up to 50 deals, says how many it left out, and notes anything it could not check. If a date range isn't valid, for example an end date before the start date, your assistant tells you what to change instead of failing with a generic error. Dates can be entered as MM/DD/YYYY.
Product gaps summary. You can limit it to Spaces with recent activity or to one owner. It groups the same gap across accounts into one row and shows its severity (the most severe among the accounts), the accounts it affects, when it was last mentioned and any interim solution. It flags gaps nobody has mentioned in more than 90 days as possibly stale, and lists the Spaces it could not read. If a Space's gaps are cut short in a large request, your assistant asks again for each of those Spaces on its own, so the summary covers every Space. If any Space is still partial, the summary says so.
Reading product gaps and deals does not use credits, does not start a conversation, and does not add anything to what Ava remembers. This is not a raw data export. There is no tool to page through call transcripts or meetings. If you want a quote from a call, ask Ava for it in your question.
To cover several accounts at once, ask your assistant for your product gaps or deals, or list your Spaces and then ask Ava the same question about each one.
Teach Ava from your assistant
See best practices here.
Trainers and Admins can teach Ava from a connected assistant, in the same way they teach it in Training in Ava. You write what Ava should learn, and can name the sources it comes from and a product line. Ava's training agent replies, and what you taught is staged to be learned.
Each message you send is one turn of a teaching session. Your assistant gets back Ava's reply, a session ID, the session's learning status, and a receipt of what is staged so far.
Learning starts on its own after a short countdown. There is no confirmation step and nothing to do in Ava afterward.
To keep teaching in the same session, your assistant sends the session ID with the next message. A session accepts more teaching for seven days after its last message. After that, start a new session.
If you name a product line, it belongs to the session. Everything taught in that session is tagged with it, and a later message can't switch the session to a different product line.
One message at a time per session. If you send a second message while the first is still running, Ava asks you to wait. Different sessions can run side by side.
Teaching is limited to 60 messages per person and 300 per organization every 10 minutes. If your assistant is told to slow down, wait a bit and try again.
Good to know: Teaching is the one exception to "Ava does not learn from a consultation". Questions you ask about a Space are still never learned. Material you teach in a teaching session is. Ava can read web pages and search the web while it teaches, and treats text in your material as information to learn, not as instructions to follow.
Ask self-contained questions
Ava does not see the conversation you are having with your assistant. Put everything Ava needs into the question itself. "What did we agree on pricing?" will not work if the account is only named earlier in your assistant's chat. Use actual dates ("between September 15 and September 22") rather than "last week".
Follow-up questions about the same Space do carry over. Ava keeps one running consultation per Space, so what you asked before, and what it answered, are still there the next time you ask about that Space.
Consultations are kept apart from your work in Ava
A consultation does not appear in the Space's threads, or in your conversation history in Ava.
Ava does not learn from a consultation. Nothing you ask here is added to what it remembers.
Ava will not change your CRM from a consultation. Asking it to update a record gets you an answer, not a change. The only thing your assistant can change in Ava is adding a Space to your list, and only when you ask for it or confirm it.
Consultations are yours alone. Someone else's consultation is never reachable from yours.
Good to know
A Space you have archived is not in the list, but you can still ask about it if you have its identifier.
One question at a time per Space. Ask a second question about the same Space while the first is still being answered and Ava asks you to wait. A question about a different Space goes ahead.
A very long answer, or a very long list of Spaces, is shortened, and the result says that it was.
Ava answers on its standard model. A model you chose inside Ava, a lens, or a call selection does not apply to a consultation.
A question that arrives without a Space, or with a Space you do not have access to, is refused rather than answered from somewhere else.
For administrators: a consultation counts the Space as active for the billing period. It does not appear in the usage dashboard's interaction counts.
Reading deals needs your Salesforce connection in Ava. If it is missing or has expired, your assistant is told to connect or reconnect Salesforce in Ava.
A very long list of gaps or deals is shortened, and the result says that it was. If Salesforce stopped returning rows before your whole pipeline was read, the result says so, so a partial list is never mistaken for the full one.
Deals reads are limited so they don't use up your organization's Salesforce allowance. If your assistant asks you to try again in a moment, wait a bit and repeat the request.
Adding a Space, reading Training knowledge, and teaching Ava each need a permission. If you connected your assistant before they were added, disconnect and connect again once, and approve the access request. Microsoft Copilot Studio admins must add the permission to the connector by hand.
A large result is trimmed to fit your assistant's size limit, and the result says that it was. A product gaps result still includes every Space you asked for.
Reading the Space list, searching accounts and the report prompts use no credits.
Training knowledge tools, including teaching, are available only to people with the Trainer or Admin role. Anyone else who asks gets a message that their role doesn't allow it.
Troubleshooting
Ava does not appear after connecting. Quit and reopen Claude.
Sign-in fails or loops. Sign in to the Ava web application first to confirm your account works, then try connecting again.
"List my Spaces" comes back empty. You have no Spaces in Ava yet. Create or join a Space in the web application first.
Ava says it is busy. A question about that Space is still being answered. Wait for it to finish, then ask again.
"Connect or reconnect Salesforce in Ava." Open Ava, go to your Salesforce integration, and reconnect, then ask your assistant again.
"This Space isn't available" or a Space is missing. If a Salesforce account has a Space in your organization but you haven't added it in Ava yet, your assistant tells you to add it in Ava. You can also ask your assistant to add it for you, then wait until the Space shows as ready before asking about it.
Your assistant can't add a Space. It tells you why. If it says to connect or reconnect Salesforce, do that in Ava and ask again. If the account belongs to a different organization, it can't be added to yours.
Learn more in About Spaces.
