When a director wants to know who owes the company the most and for how long, someone has to open the invoice list in Flexi, set a filter, sort it and export the result to a spreadsheet. With Claude and an MCP server you just ask, and the answer comes straight from your data. This article covers what you need, how to set it up and where the limits are.<\/p>\n
You need three things: a Claude account, ABRA Flexi with an active API, and an MCP server in between.<\/strong> It works with Flexi in the cloud and on your own server. Connecting in Claude takes a few minutes. Read-only is the default. Writing is switched on separately, and you confirm every action.<\/strong><\/p>\n<\/div>\n Most companies run ABRA Flexi in the vendor’s cloud, where the only way to the data is the REST API. Orders from online stores, CRM data and bank statements reach Flexi the same way. So the MCP server uses the API too. That means it works the same with cloud Flexi and with a self-hosted installation, nothing is installed on your side, and what the AI can see is set by the API user’s permissions in Flexi. If you run Flexi on your own server, the API has to be reachable from the internet; access can be limited to our server’s IP address. We agree the details with your IT team.<\/p>\n Since 1 January 2025 the Flexi API is paid and works only with the API Access<\/em> add-on (vendor’s help centre<\/a>). The add-on price and the daily API request limit for your plan are in the ABRA Flexi price list<\/a>. For AI over Flexi this means three things:<\/p>\n In Claude Code you add the server with one command: Claude doesn’t just find the data, it works with it straight away: it drafts a payment reminder, builds a table or points out an unusual swing. On the higher plan it can also write, for example prepare an invoice, an order or a new contact. It saves them only after you confirm.<\/p>\n AI over your ERP won’t replace integration. If orders from Shopify need to land in Flexi automatically every few minutes, even at three in the morning, you need classic system integration<\/a>. We cover the options in our article on connecting Shopify to an ERP<\/a>. MCP connects people with data, integration connects systems with each other. Most companies end up using both.<\/p>\n<\/div>\n You don’t have to choose, the same MCP server works with both. The MCP<\/a> standard was created by Anthropic, so Claude has supported it the longest. In ChatGPT you add a custom MCP server in developer mode, available on the Business, Enterprise and Edu plans. If your company already has one subscription, start with that. The same steps for other systems are in our guides How to connect Claude to POHODA<\/a> and to Helios iNuvio<\/a>.<\/p>\n We go into more detail in AI security over company data<\/a>. How we handle it is described on our security model<\/a> page.<\/p>\n Yes, that is the most common case. The MCP server connects to Flexi through the REST API, just like an online store or any other app.<\/p>\n<\/details>\n It works the same way, the API just has to be reachable from the internet. We agree the details with your IT team.<\/p>\n<\/details>\n If you don’t have it yet, yes. Without it the Flexi API doesn’t work. If your Flexi is connected to an online store or a CRM, you are probably paying for it already and only need to create another API user for AI.<\/p>\n<\/details>\n It shouldn’t. Each query loads only a limited number of rows and only the columns it needs, so a typical query puts less load on Flexi than a manual export. We set the limits during implementation.<\/p>\n<\/details>\n No. The AI sees only the modules you allow, and never more than the API user’s permissions permit. Sensitive areas, payroll above all, can be excluded entirely.<\/p>\n<\/details>\n No. You name the company in your question, and every answer states which company the data comes from.<\/p>\n<\/details>\n Claude receives only the data the MCP server returns for a specific question, never the whole of Flexi. On the Team and Enterprise plans conversations are not used to train models. On Free, Pro and Max each user decides this in the privacy settings. Our MCP server runs in the EU and stores data only in the audit log.<\/p>\n<\/details>\n You pay for Claude or ChatGPT and for the API Access add-on directly to their providers. With us you pay for data access: from \u20ac99 per month excl. VAT per company for read-only, \u20ac199 with write access, plus a one-off implementation fee from \u20ac800 covering access setup, module configuration and training. Details are on the ABRA Flexi MCP<\/a> page.<\/p>\n<\/details>\n","protected":false},"excerpt":{"rendered":" You can ask Claude about your ABRA Flexi data in plain language. All it takes is an active API and an MCP server, in the cloud or on your own server. What you need, how to set it up and where the limits are.<\/p>\n","protected":false},"author":0,"featured_media":72155,"template":"","erpedie-topic":[98,104],"class_list":["post-72162","erpedie","type-erpedie","status-publish","has-post-thumbnail","hentry"],"acf":[],"_links":{"self":[{"href":"https:\/\/symmy.com\/en\/wp-json\/wp\/v2\/erpedie\/72162","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/symmy.com\/en\/wp-json\/wp\/v2\/erpedie"}],"about":[{"href":"https:\/\/symmy.com\/en\/wp-json\/wp\/v2\/types\/erpedie"}],"version-history":[{"count":0,"href":"https:\/\/symmy.com\/en\/wp-json\/wp\/v2\/erpedie\/72162\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/symmy.com\/en\/wp-json\/wp\/v2\/media\/72155"}],"wp:attachment":[{"href":"https:\/\/symmy.com\/en\/wp-json\/wp\/v2\/media?parent=72162"}],"wp:term":[{"taxonomy":"erpedie-topic","embeddable":true,"href":"https:\/\/symmy.com\/en\/wp-json\/wp\/v2\/erpedie-topic?post=72162"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}Why the REST API<\/h2>\n
You need the API Access add-on<\/h2>\n
\n
What you will need<\/h2>\n
\n
Connecting in three steps<\/h2>\n
\n
https:\/\/mcp.symmy.ai\/abra-flexi<\/code>) and sign in. You never enter your Flexi password into Claude, and you can revoke access at any time.<\/li>\nclaude mcp add --transport http abra-flexi https:\/\/mcp.symmy.ai\/abra-flexi<\/code>.<\/p>\nWhat you can ask<\/h2>\n
\n
Where it won’t help<\/h2>\n
Claude or ChatGPT?<\/h2>\n
Security: four questions before you switch it on<\/h2>\n
\n
Frequently asked questions<\/h2>\n
Our ABRA Flexi is in the cloud. Will it work?<\/summary>\n
What if we run Flexi on our own server?<\/summary>\n
Do we have to buy the API Access add-on?<\/summary>\n
Won’t it slow Flexi down for other users?<\/summary>\n
Will the AI see all of our data in Flexi?<\/summary>\n
We run several companies in Flexi. Is that a problem?<\/summary>\n
Will Anthropic see our accounts?<\/summary>\n
How much does it cost?<\/summary>\n