Connect Vara to AI Assistants

Once connected, you can talk an experiment through in the AI you already use, then ask it to save the experiment to Vara and put it on your plan.

Before you start

You need a Vara account. If you don't have one, you will be asked to sign in while connecting, and you can sign up from that page.
Every client uses the same connection URL:

https://varaos.com/api/mcp

Any Vara account can connect; there is nothing extra to enable and connecting is free. Usage counts the same way as in the web app (see Common questions below). Pick the section for the AI you use and follow it; each takes about two minutes. Menu names change as clients are updated, so follow what your client currently shows.

Step 1: add Vara to your AI

Claude (web and desktop)

In Claude, open Settings → Connectors and click Add custom connector.
Enter Vara as the name, paste the connection URL above, and save.
Click Connect. Your browser opens the Vara sign-in and authorization page; sign in and click Allow.
Back in Claude, start a new chat and try "List my Vara experiments".

If Claude does not use Vara, check that Vara is turned on in the tools menu of the message box (its location may vary).

ChatGPT

Vara is not listed in the ChatGPT plugin directory yet, so for now you add it by hand as a custom MCP server. According to OpenAI's help center, you now add it directly from Plugins and do not need to turn on developer mode first. It can only be added in ChatGPT on the web, not on mobile.

Whether you can add it, and whether it can write, depends on what ChatGPT shows you. If you are not sure it can write once connected, use the connection check under Common questions below.

In ChatGPT on the web, open Plugins in the sidebar, click the plus button and choose Add custom MCP server. Menu names may differ in the current ChatGPT interface.
Enter Vara as the name; the description is optional. Under Connection, enter the connection URL above, https://varaos.com/api/mcp.
Choose OAuth for authentication. Read the risk warning, tick I understand and want to continue, click Create as a plugin and check the Vara tools it lists.
Install the plugin when prompted. When your browser opens the Vara sign-in and authorization page, sign in and click Allow.
Start a new chat, type @ in the message box and select Vara, then try "List my Vara experiments".

ChatGPT may ask you to confirm before it writes data; confirm when prompted.

Other clients (WorkBuddy, Doubao and similar)

Where to add a custom MCP server, and how it signs in, differs between clients. Look in Settings for a page called MCP, Connectors or Tools, and a button such as Add custom MCP or Add remote server; follow the names your client shows. The client will use one of these two methods.

Method 1: the client opens a browser sign-in

Enter the connection URL above as the server URL.
If you can pick a transport, choose Streamable HTTP (some clients call it HTTP).
Save, open the browser when prompted, sign in to Vara and click Allow.

Method 2: the client only takes a URL and request headers

Create an API key in Settings → API and save it (it is shown only once). Choose read and write if the assistant should change data (create experiments, log runs, edit the plan and so on); read-only is the default.
Set the server URL to https://varaos.com/api/mcp and the transport to Streamable HTTP.
Add a request header named Authorization whose value is Bearer, a space, then your key:

Authorization: Bearer vara_sk_...

This skips the authorization page: the key itself is the credential. No API access is needed, and usage rules are the same as for an authorized connection.
Key-based connections do not appear under Connected apps. To disconnect, revoke the key in Settings → API; it stops working immediately.
Treat the key like a password: do not share it, paste it into a chat or include it in screenshots. If you think it leaked, revoke it and create a new one.

Advanced: Claude Code

Run this in your terminal:

claude mcp add --transport http vara https://varaos.com/api/mcp

Open Claude Code, type /mcp and select vara to authorize. Your browser opens the Vara sign-in and authorization page; click Allow and return to the terminal.

Advanced: Cursor

On a computer with Cursor installed, click this button:

Add to Cursor

If the button does nothing, add this to mcp.json yourself and save:

{
  "mcpServers": {
    "vara": { "url": "https://varaos.com/api/mcp" }
  }
}

Advanced: Codex

Add this to ~/.codex/config.toml and save:

[mcp_servers.vara]
url = "https://varaos.com/api/mcp"

Run this in your terminal. Your browser opens the Vara sign-in and authorization page; click Allow:

codex mcp login vara

Quit Codex completely, reopen it, start a new session and try "List my Vara experiments".

The ChatGPT desktop app, the Codex CLI and the IDE extension share this configuration, so one setup covers all of them. If you connected Vara before with an API key and the local server, remove the old [mcp_servers.vara] block and use the configuration above.

Any other client that supports remote MCP with OAuth sign-in can connect with the same URL. For clients that can only run a local command, see the local setup in the MCP reference (the local server requires API access).

Step 2: what to say once connected

Talk normally and mention "Vara" so the assistant knows to use it. You can copy the requests below and swap in your own experiment names and dates. If it is not clear which experiment you mean, the assistant lists candidates for you to pick from instead of creating a new one.

Save the chat as an experiment

Turn the experiment we just discussed into a Vara experiment. Show me the name and steps before saving.

Add a step

Add a step to my "SFRP1 qPCR" experiment in Vara: incubate at 37°C for 30 minutes, and put the primer concentration in the step notes.

Plan next week

Plan next week's bench work in Vara: passage cells on Monday, Western blot on Wednesday, RNA extraction on Friday. Show me the plan before saving.

Move a plan item

Move Wednesday's Western blot in my Vara plan to Thursday at 2 pm.

Remove a plan item

Remove tomorrow's "Order reagents" item from my Vara plan.

Log a run

Log today's run of "SFRP1 qPCR" in Vara: amplification curves look normal, I forgot the template in well 3. My conclusion: SFRP1 is downregulated.

List recent experiments

List the 5 experiments I updated most recently in Vara.

Today's plan

What is on my Vara plan for today?

Shortcuts

The Vara connector also comes with a few shortcuts (MCP prompts). Picking one sends a ready-made request to the assistant. Clients that support shortcuts list them: in Claude Code, type / to see them, for example /mcp__vara__plan_my_week; in Claude on the web and desktop they are usually in the + menu of the message box (its location may vary). In clients without shortcuts, just use the requests above.

save_conversation_as_experiment: Turns the current chat into a Vara experiment and asks you to confirm the name and steps before saving. Optional: field.
plan_my_week: Lists the week's plan first, then adds, moves or removes items after you agree. Optional: start date.
log_todays_run: Finds the experiment and asks you to confirm it, then records today's run in your own words. Optional: experiment name.
today_in_vara: Shows today's plan items and your recently updated experiments. Read only.

Rather not set anything up?

You can skip connecting and use Import from external AI instead: copy a prompt from Vara, paste it into any AI, then paste the result back into Vara, review it and save it as an experiment.

After signing in, open Import from external AI (also in the New menu in the sidebar and on the home page) and copy the prompt.
Paste the prompt together with your paper, notes or idea into ChatGPT, Claude, DeepSeek or another AI and let it generate the result.
Paste or upload the result on the same page. Vara parses it into a review page, and nothing is saved until you confirm.

See Import & Parse for details. This needs a copy and paste each time and can't plan or log runs, so connecting is easier if you use it often.

Access: read-only or read and write

The authorization page shows the app asking for access and the address it returns to after you approve. You choose the access level:

Read-only: view your experiments, sub-experiments, run logs, plans, references and materials.
Read and write: everything in read-only, plus create and edit these records, upload attachments, and delete plan items, attachments and reference links.
With either level, core data (experiments, sub-experiments and run logs) cannot be deleted.
When the AI edits an existing experiment, sub-experiment or run, the edit waits in Pending changes by default and takes effect after you accept it in Vara. See the MCP reference.
A connection cannot read your AI provider keys, account settings or password.
The access token the client receives can also call the Vara /api/v1 REST API directly. Like an API key, direct REST calls require API access (a trial or a subscription), with the same scope and rate limits. The LLM gateway (/v1/chat/completions, /v1/models) does not accept it.

If the app name or return address on the authorization page is not the one you just started connecting, click Deny.

Common questions

Where do I revoke access?

In Settings → API, find the app under Connected apps and click Disconnect. This takes effect immediately: the app can no longer access your Vara data and must be authorized again to reconnect. Removing the connector only in your AI client does not cancel the authorization on the Vara side right away, so disconnect in Settings as well. For a key-based connection, revoke the key.

What are the Free plan limits?

Connecting is free, and usage counts the same way as in the web app: the Free plan allows up to 3 experiments, and actions that use Vara's own AI (such as reading reference details from a PDF, or importing experiment notes that are not in the import format) count toward your AI quota. When the experiment limit is reached, delete one you no longer need or upgrade; the AI quota refills automatically over time. The assistant gets a message explaining the limit. Plans are on the Pricing page, and you don't need to reconnect after upgrading.

Will the AI platform train on my data?

What Vara can say for certain: Vara only reads and writes within the access you granted, and only sends data to the client you authorized. Vara does not receive your conversations with the AI, only the individual operations it sends to complete your request.

When you ask the assistant to do something, the Vara data its tools return (for example experiment names, steps, parameters and run notes) becomes part of your conversation with that AI. Whether that platform uses it for training, and how long it keeps it, depends on the platform's policies and your settings there; check the platform's own documentation. Content already in a conversation has to be deleted on that platform. See the Privacy Policy for details.

I said "save it to Vara" and nothing happened

Check that the Vara connector is turned on for this chat, and mention "Vara" explicitly. If the assistant says it has no such tool, check in your client that Vara shows as connected, and reconnect if it doesn't.

Connection check: not sure it is connected, or the assistant says it can't do something

Copy the text below to your assistant. It only creates one test plan item and removes it at the end, so your existing data is not touched. If it still does not work, send the assistant's full reply to Support and we can tell which step went wrong.

Check whether my Vara connection can read and write: first list the Vara tools you can use; then add a "Connection test" item to tomorrow in my Vara plan, mark it done, and finally remove it. For each step, tell me which tool you called and what it returned; if a step does not work, give me the reason or error exactly as it is, and do not skip it.

Can the assistant upload files?

A remote connection cannot read files on your computer. When the assistant uploads an attachment, it can pass a public file link (up to 25MB) or upload a small file directly (up to 3MB). Drag larger local files into the Vara web app instead.

It won't connect, or keeps showing a connection error

Check the URL first. It must be https://varaos.com/api/mcp, without www and with /api/mcp at the end. Sign in with your own Vara account on the authorization page. If you are not sent back to the client after signing in, remove the connector in the client and add it again.

It says the API key is invalid or missing, but the key is fine

Most likely the URL has www in it. Requests to www.varaapp.site are redirected to varaapp.site, and the key is dropped on the way, so Vara never receives it. Remove the www. from the URL (for the local server, https://varaos.com/api/v1), then quit your client completely and reopen it. Clients read their configuration only at startup and keep using the old URL until they restart.

It says access is not allowed, or that the connection is read-only

The connection was authorized as read-only. To write data, disconnect the app in Settings → API, connect again from the client, and leave read-only unticked on the authorization page. If you connected with an API key, the key was created read-only: create a read and write key and put it in the request header instead.

Network errors, timeouts, or a security check page during sign-in

The Vara site runs behind Cloudflare, so a security check page may appear before sign-in; complete it and authorization continues. For an occasional network error or timeout, wait a few minutes and retry. Some company or campus networks and proxies block these connections, so try another network. Claude, ChatGPT and some other AI platforms are not available in every region, and Vara cannot work around that.

It says the connection is no longer authorized

This appears after the connection is disconnected or its authorization expires. Reconnect Vara in your client. Still stuck? See Support.