Introducing Basedash MCP connectors

Today we’re launching MCP connectors. They let you plug any external Model Context Protocol server into Basedash so the agent can take action on the data it already reads.

Basedash already pulls from your databases, warehouses, and 750+ SaaS tools. Connect an MCP server (Linear, HubSpot, Slack, Resend, Notion, GitHub, or your own internal one), and every tool that server exposes becomes available to the Basedash agent inside chat and Automations.

You can ask it to email this week’s signups a personalized welcome based on the features they set up, file a Linear bug from a support ticket and link the user record, or update HubSpot leads for everyone who hit the paywall yesterday.

Why we built MCP connectors

BI tools have always been read-only. They can tell you a signup cohort is stuck, a support issue is recurring, or a customer record needs attention, but the moment you want to do something about it, you leave the dashboard and open another tab.

That hand-off is where most “data-driven” workflows break down. The analysis happens in one place and the action in another, and the answer to “what should we do about this?” gets lost on the way to the inbox, the CRM, or the issue tracker.

Once you connect an external MCP server, its tools sit next to your data sources. The agent can read from your database and act in Linear in the same turn, or compare cohorts and update HubSpot from the same prompt, so the follow-through happens where the analysis happens.

This is the natural next step for the agent. We’ve spent the last year making it a great analyst, and now it can also be a great operator, with all the same governance.

How MCP connectors work

Adding a connector takes about a minute:

1. Open the command menu in Basedash, go to Data sources, and choose Add MCP server.
2. Enter a name and the remote MCP server URL. Streamable HTTP and SSE are both supported. Add headers if the server needs them.
3. Complete OAuth if the server prompts for it. Basedash syncs the available tools.
4. Use it. The connector now appears alongside your other data sources in the command menu, and every synced tool is available to the agent inside chat and Automations.

Behind the scenes, Basedash speaks remote MCP over streamable HTTP or SSE, handles OAuth refresh, and re-syncs tools whenever the upstream server changes.

If it speaks MCP, it works

Our customers are connecting the apps they already work in.

• Linear: file bugs, update issues, link the right user record
• HubSpot: update leads, sync deals, fire workflows from product signal
• Slack: post into the right channel based on what just happened in the data
• Resend: send personalized email to a list the agent just generated
• Notion: drop a structured brief into the right doc or database
• GitHub: open an investigation issue with the chart and the SQL attached
• Stripe, Intercom, your own internal MCP server, anything that speaks remote MCP: same flow, same governance

If a service exposes a remote MCP server, Basedash can use it. If you have an internal MCP server in front of an internal API, point Basedash at that too.

What you can ask for

The most common pattern is to read your governed data, then act in another app.

• “Email this week’s signups a personalized welcome based on the features they actually set up.” Reads your production DB and product events; sends through resend.send_email.
• “File a Linear bug from this support ticket and link the user record.” Reads Intercom and your production DB; creates the issue through linear.create_issue.
• “Update the HubSpot lead for everyone who hit the paywall this week.” Reads your product events; updates contacts through hubspot.update_lead.
• “Post a Slack thread in #wins for any 100+ seat company that signed up today, with the account context.” Reads your CRM and product DB; posts via slack.post_message.
• “Open a GitHub investigation issue with this dashboard attached when error rate doubles.” Reads your logs and alerts; opens the issue via github.create_issue.

Each one pairs a governed read with a governed action in a single prompt.

Trust the tools you’ve vetted. Gate the rest.

Read-only BI is safe by default, but taking actions isn’t, so we built MCP connectors around explicit human review.

Every synced tool gets one of three access modes:

• Always allow. Trusted tools run without interruption: read-only lookups, well-scoped helpers, anything you’ve vetted.
• Needs approval. Higher-stakes actions pause for a human review of the full payload before they run. New tools start here by default.
• Blocked. For anything you don’t want the agent to touch: flip a tool off and it disappears from the available toolset.

When the agent calls a Needs approval tool, a review card appears in the chat with the exact payload (recipients, subject, body, every field) so you can approve or reject it before anything fires. The same approval flow works in Slack, so the team can keep moving without jumping back into Basedash.

Scope each connector to the right people

Connectors also carry their own audience. You can:

• Open a connector up to everyone in the org
• Restrict it to specific groups, such as giving only Support access to Intercom and only Growth access to HubSpot
• Hand it to specific members when the action is sensitive or scoped to a role

Pair connector-level scoping with tool-level approval modes and you get fine-grained control over both who can use a connector and which actions inside it run automatically.

Run the same flow on a schedule

Every tool a connector exposes is also available to Automations. The one-off agent run that emails this week’s signups becomes a weekday-9am workflow that sends those emails every week, with the same prompt, the same governance, and no extra setup.

A one-prompt experiment can become a recurring operation as soon as it works: a daily ICP welcome, a weekly HubSpot refresh, real-time bug filing from support escalations, or sales pipeline updates the moment a deal hits a stage.

How we use MCP connectors at Basedash

We’ve been dogfooding connectors internally. A few favorites:

• Welcome emails: every weekday morning, an automation reads new signups from production, joins them with what they did in product, and sends a personalized welcome through Resend. After the initial approval, it runs without a human in the loop.
• Bug intake: when a support ticket comes in that looks like a bug, the agent files a Linear issue with the customer context attached, then posts a link back in Intercom.
• Customer-health flags: every Monday, accounts trending down get pushed into HubSpot as tasks for the right owner, alongside the chart that flagged them.

These hand-offs between teams used to be slow. Now the agent runs them, and the team reviews the actions that matter most.

Getting started

MCP connectors are available today for all Basedash workspaces.

1. Sign up for Basedash (or log in)
2. Open the command menu and go to Data sources → Add MCP server
3. Paste a remote MCP server URL, complete OAuth if prompted, and start using its tools in chat
4. Pair with Automations when you’re ready to run the same flow on a schedule

For more, see the MCP connectors feature page or the docs.

What’s next

The Basedash MCP server lets any AI client read your governed data, and MCP connectors let the Basedash agent take action in any tool your team already uses. With both, plus AI chat, Insights, Automations, and the Dashboard Agent, governed analytics and AI-driven action now live in the same Basedash workspace.

Try MCP connectors today by connecting the first app you want the agent to act in.