Your Fondaro pipeline, inside the AI tools you already use
If Codex CLI or Claude Code is already where you plan the day, write copy or work through a difficult client request, your CRM should not become a separate research project. Fondaro MCP connects those tools to your real pipeline and property sources. You can ask a question in plain language, let the assistant do the reading, and keep working where you already are.
Not from a pasted spreadsheet. Not from an export that was stale before you opened it. From the same Fondaro record your team uses on the sales floor.
MCP stands for Model Context Protocol. It is an open standard that lets an AI client use approved tools from another product. Fondaro runs the server. You choose the client, the permissions and when access ends.
What you can ask it to do
Picture a normal morning in a Marbella agency. A Swedish buyer is due a follow-up, two viewings need tasks, yesterday's calls need reviewing, and a Dutch client wants a fresh shortlist around Estepona.
With Fondaro MCP connected, you can ask your assistant to:
- Find a lead and summarise the real contact record and activity timeline.
- Read stored call transcripts and the existing call analysis before you prepare the next conversation.
- Add a note, create a task or move a lead to the right pipeline stage.
- Create and update deals, then close them as won or lost when the outcome is known.
- Search your internal listings, Resales Online, Zoddak, Inmobalia and 19 public property portals for a buyer.
The assistant can turn those facts into a draft or a plan in its own workspace. Fondaro supplies the grounded record and the tools to work it. The public connection does not send emails or place calls.
Same person, same permissions
Connecting an AI tool should not create a side door into the agency.
Every Fondaro MCP request acts as the person who created the connection. A team member can reach only the leads assigned to them. An organisation admin can work across the organisation. Admin-only actions, such as changing lead ownership across the team, stay hidden from members.
Each connection also gets its own key. You decide whether that key can read CRM data, change CRM data, search properties, or use a combination of the three. You can set an expiry, create a replacement and revoke the old key at any time. Remove someone from the organisation and their connection stops working too.
Start with the smallest permission set that does the job. A client used for morning pipeline summaries may need CRM read access and nothing else. Add write access only when you want it to change records.
Why we chose a key instead of a browser promise
The easy launch story would have been: paste a URL, sign in, done.
We tested that route. The browser flow worked, but the access token did not give our server the resource-level proof we required. So we stopped. A connection that looks effortless is not worth shipping if the boundary underneath it is vague.
Fondaro MCP therefore launches with scoped, revocable keys for clients that can keep a private authorisation value. The key is shown once, stored by Fondaro only as a secure hash, and never needs to sit in a project file. The dashboard guides you through loading it privately on your computer and verifies when Fondaro receives the first request.
This means hosted ChatGPT apps and hosted Claude connectors are not supported today. Codex CLI and Claude Code have guided setup, Cursor has an advanced manual path, and other clients work when they can send a private authorisation header. We will not pretend every AI product supports the same safe connection model.
From Settings to the first request
The connection wizard lives under Settings, Integrations, then Fondaro MCP. It keeps the job to four steps:
- Choose Codex CLI, Claude Code, Cursor or another compatible client.
- Create a dedicated key and choose what it can access.
- Save the key in your password manager and load it privately for the current terminal session.
- Configure the client, make one request and let Fondaro verify that it arrived.
The full key never goes into the saved client configuration. The configuration holds the name of a private value instead. If you lose the key, Fondaro cannot reveal it again. Create a replacement, confirm it works, then revoke the old one.
Autopilot inside Fondaro, MCP outside it
Autopilot remains the assistant built into Fondaro. It reads the same pipeline, drafts the next move and gives you a review step inside the dashboard.
Fondaro MCP opens the other direction. It lets the AI workspace you chose reach Fondaro with a carefully limited set of tools. Use Autopilot when you want the native sales-floor experience. Use MCP when Codex CLI, Claude Code or Cursor is already the place where the work is taking shape. Use both if that fits your day. They meet in the same live CRM, so a note created in one is there for the other.
The point is not to add another AI tab. It is to remove one more reason for your pipeline to fall out of the work.
Bring your real pipeline into the AI workspace you already use.