Viewert in Claude and ChatGPT
Let your AI assistant read your Librams, save decisions back to them and draw your systems as Graphs, with only the permissions you approve.
Connect Viewert once and your assistant works from your own documents instead of starting from nothing each time. It acts on your account only, and only as far as you allow.
Read
List your Librams, read one as Markdown, measure its size and search inside it. The assistant asks for what it needs rather than loading everything.
Save
Save a fact or a decision to a Libram so it is there next time. Shortening or combining Vellums is offered as a proposal; nothing is replaced until you agree, and the earlier text stays in the Vellum's history.
Draw
Turn systems you describe into a Graph (an architecture diagram) in your account, checked first for mistakes and for credentials, which are never saved.
In the Claude apps on the web, desktop and mobile, Viewert is added as a connector. Where Viewert is listed in the connector directory, add it from there; otherwise add it by its address.
Open the connector settings
In Claude, open Customize → Connectors, choose + Add, then Add custom connector. On a Team or Enterprise plan an owner adds it first, under Organization settings → Connectors (Add, then Custom, then Web); members then find it in Customize → Connectors and choose Connect.
Enter the address
Use: https://api.viewert.com/mcp
Approve on Viewert
Claude opens a Viewert page that lists every permission it asks for. Sign in if asked, read the list and choose Allow access. Choose Deny and nothing is connected.
Try it
"What Librams do I have in Viewert?" Claude lists them and asks which to use.
In a terminal, add Viewert as an HTTP server and sign in when Claude Code opens the approval page:
Add it
claude mcp add --transport http viewert https://api.viewert.com/mcp
Sign in
Run /mcp inside Claude Code, choose viewert and follow the link to approve it.
ChatGPT adds Viewert as an app that uses the same address. Menu names change from time to time; look for the place that adds your own app or connector. Saving anything to Viewert, a Memory included, is a write action, and OpenAI's own pages differ about which ChatGPT plans allow write actions from your own app. After connecting, ask it to save one test Memory (the Memories recipe has the exact words) to find out for your plan.
Open the apps settings
In ChatGPT, open Settings → Apps (older versions call it Connectors). Where Viewert is listed in the plugin directory, choose it there. Otherwise turn on developer mode and create a new app.
Enter the address
Server URL: https://api.viewert.com/mcp Authentication: OAuth. Viewert registers the app itself, so no client id or secret is needed.
Approve on Viewert
The same permission page opens. Allow access, then start a chat and choose Viewert from the tools menu.
Each permission is listed on the approval page, and you can see what you approved later.
- What it lets the assistant do
- List Librams, read them, search them
- What it lets the assistant do
- Read a Vellum to shorten it or combine it with others
- What it lets the assistant do
- Save to a Libram, and save a shortened or combined Vellum once you agree
- What it lets the assistant do
- Recall the short notes you keep for your AIs, and look up where one came from
- What it lets the assistant do
- Suggest new Memories, replacements and removals. They wait for your review until you choose to trust the app; an app can never edit or delete a Memory
- What it lets the assistant do
- List your Graphs and check a diagram before saving it
- What it lets the assistant do
- Save a Graph. Always private: an assistant can never publish or share one
- What it lets the assistant do
- Only for apps that sign you in. Claude does not ask for them; ChatGPT may, to recognise your account
| Permission | What it lets the assistant do |
|---|---|
| Read your Librams | List Librams, read them, search them |
| Read your Vellum documents | Read a Vellum to shorten it or combine it with others |
| Create and update your Librams / Vellums | Save to a Libram, and save a shortened or combined Vellum once you agree |
| Read your Memories | Recall the short notes you keep for your AIs, and look up where one came from |
| Propose Memories | Suggest new Memories, replacements and removals. They wait for your review until you choose to trust the app; an app can never edit or delete a Memory |
| See your Graphs | List your Graphs and check a diagram before saving it |
| Create, change and delete your Graphs | Save a Graph. Always private: an assistant can never publish or share one |
| Know that it is you / see your email address | Only for apps that sign you in. Claude does not ask for them; ChatGPT may, to recognise your account |
Most problems are the address or a missing permission.
The approval page says the app is unknown
Remove the connector and add it again; an assistant registers itself the first time it connects.
The assistant says it cannot reach Viewert
Check the address is exactly https://api.viewert.com/mcp, with no extra path.
A tool says it needs a permission
Revoke the connection and connect again, approving the permission it names.
Your plan's limit is reached
Saving a Graph counts against your plan's Graph limit, and the assistant is told so. See Plans, Teams & Billing.
Still stuck
Write to us from the Help page and tell us which assistant you used.