Short answer: if your tool has no MCP server but does connect to Make, you can still drive it from Claude or ChatGPT. You build a small Make scenario for each job (“add a contact”, “create a post”), switch it to on-demand, and connect Make’s own MCP server to your AI. The AI then sees each scenario as a tool it can call. It works on Make’s free plan.
When this is the right move
Look at your tool page on ToolsAir. A bridge makes sense when the status is No MCP, Promised with no firm date, or Higher tier and you do not want to upgrade.
It also helps when you own several small tools. One Make connection can expose jobs from all of them, so your AI has one place to go instead of five half-finished integrations.
It is not worth it when the tool already ships a remote MCP server on your tier. Use that directly.
How the bridge works
There are three pieces:
- Your tool, connected to Make through Make’s app for it, its API or its webhooks.
- A Make scenario for each job. It takes a few inputs (a name, an email, a URL), does the work in your tool and returns a result.
- Make’s MCP server. Make hosts it in the cloud. It shows your on-demand scenarios to the AI as tools.
From the AI’s side, “run the scenario called Add lead to CRM” looks the same as calling a native MCP tool.
Before you start
Check one thing first: can your tool talk to Make at all?
- Search for it in Make’s app list. If it is there, you are set.
- No app? Check whether the tool has an API. Make’s HTTP module can call almost any API with a key.
- No API either? See whether it can send webhooks, for example “on new order”. Make can receive those, but that only covers events coming out of the tool, not commands going in.
If the answer to all three is no, a bridge will not help. Ask the vendor for an API or MCP, and get a date.
Step by step
1. Build one small scenario. Start with the job you do most often. Keep it to a single task: “create a contact”, not “manage my CRM”.
2. Add inputs and outputs. In the scenario settings, define what the AI must give the scenario (for example email and name) and what the scenario returns (for example contact_id and a short status). Clear names and one-line descriptions matter: they are what the AI reads to decide when to use the tool.
3. Make it available. Activate the scenario and set its schedule to on demand. Only active, on-demand scenarios show up as MCP tools.
4. Connect Make to your AI. In Claude, add a custom connector with Make’s MCP server address and sign in with OAuth. Make’s help center has the exact address for your region and a short walkthrough with Claude. ChatGPT works too, through Developer Mode, within the limits of your ChatGPT plan.
5. Choose permissions. When you connect, Make asks which scopes to grant. Give only what you need. OAuth also lets you limit access to one Make organization, which an MCP token does not.
6. Test it in chat. Ask something like “Which tools do you have from Make?” and then give it a real job. Claude will ask permission before it runs a tool the first time.
What it costs
Every scenario run uses Make credits. A simple three-step scenario costs a few credits per run. The free plan’s 1,000 credits a month covers light personal use. Credits do not roll over, so unused ones are lost at the end of the month.
If you use the bridge every day, a paid Make plan or a self-hosted n8n server will be cheaper per call. Do the maths with your real volume before you commit.
Limits to know about
- Speed. Each call runs a scenario, so expect a few seconds of delay.
- Only what you build. The AI can do exactly the jobs you made scenarios for, nothing more. That is also a safety feature.
- Read-heavy work is clumsy. “Search my last 500 records and find patterns” is better served by a real MCP server or an export.
- You maintain it. If the tool changes its API, the scenario may break. Native MCP from the vendor is still the better long-term answer.
A safer way to start
Begin with read-only scenarios, such as “find a contact” or “get today’s stats”. Add write actions once you trust the setup. For anything that sends email, charges money or publishes in public, add a final step in the scenario that waits for your approval, or keep a human in the loop in chat.