About Make
Make Skills is a collection of guidance for AI coding agents working with Make. It covers scenario management through Make’s MCP server, direct app actions with Endpoints, and secure credential requests.
Key features
- 01Explore Make accounts and scenarios
- 02Build and edit scenarios
- 03Run and debug executions
- 04Call app actions with Endpoints SDK or CLI
- 05Request credentials without exposing secrets
Use cases
- 01Understand an existing scenario and its connections
- 02Create or troubleshoot scenario workflows
- 03Run app actions through Endpoints
- 04Provision credentials for app testing
MCP support
Scenario skills use Make’s scenario-management MCP server to explore, build, operate, and debug scenarios. Endpoints actions and credential requests use separate interfaces.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://mcp.make.com/v2claude mcp add --transport http integromat-make https://mcp.make.com/v2{"mcpServers":{"integromat-make":{"type":"http","url":"https://mcp.make.com/v2"}}}Open in Cursor{"servers":{"integromat-make":{"type":"http","url":"https://mcp.make.com/v2"}}}Open in VS Code{"mcpServers":{"integromat-make":{"type":"http","url":"https://mcp.make.com/v2"}}}{"mcpServers":{"integromat-make":{"serverUrl":"https://mcp.make.com/v2"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
Tools
Details
From the project READMEMake Skills for AI Coding Agents
Give your AI coding agent deep Make expertise — for building, explaining, running and debugging Make scenarios through Make's scenario-management MCP server, calling app Endpoints through the standalone Endpoints SDK or CLI, and securely requesting credentials for Endpoints and custom-app testing. Works with agents that read skill folders and can use the relevant interface — Claude Code, Codex, Cursor, GitHub Copilot, Windsurf, Cline, and others.
Skills
| Skill | What it does |
|---|---|
| make-scenario-reference | The conventions every tool shares — scopes, the data/remark contract, structure vs configuration, the one-save write model, and what the surface refuses to author — for when a tool answers 403 or refuses something |
| make-scenario-explore | Orienting in an account — organizations, teams, listing scenarios, explaining what one does, connection health, account-wide checks |
| make-scenario-building | Creating and editing scenarios — app and module discovery, connections, resolving account-dependent values, flow control, error handling, subscenarios, AI agents; with references and complete examples |
| make-scenario-operations | Running, activating and deactivating, reviewing and debugging executions, investigating a webhook |
| make-endpoints | Direct app actions through @makehq/endpoints-sdk or its CLI — discovery, managed connections, typed calls, complete retrieval, and approval-bound writes with readback. Endpoints are in closed beta. |
| make-credential-requests | Credential provisioning for agents using Endpoints or testing custom apps — the owner receives an authorization link and supplies secrets in Make; the agent resumes with managed connection/key IDs, not raw credentials. No scenario required. |
Creating a connection in the scenario editor remains the standard manual path. Credential Requests are the recommended secure handoff for agents; they complement that editor workflow rather than replace it.
Prerequisites
- A Make account
- For scenario skills: access to the Make scenario-management MCP server.
- For
make-endpoints: an approved Node.js/terminal environment, the Endpoints package, and organization access to the Endpoints beta. See setup and interfaces. Installing this skill does not install the SDK or enable the beta. - For
make-credential-requests: access to a chosen, authorized credential-request interface, with the required Make permissions and organization/team access. Verify that interface's current contract before making a request; installing a skill or plugin does not grant access. SDK/REST requests follow their own permission contracts, not the scenario MCP server's all-or-nothing OAuth bundle.
Installation
Install the skills and configure the interface they use. Scenario skills use the MCP server; make-endpoints uses @makehq/endpoints-sdk, which includes both the SDK and make-endpoints-cli. make-credential-requests uses the chosen authorized request interface to provision credentials, not execute app actions. Neither Endpoints execution nor credential provisioning requires building a scenario. Endpoints-only use does not require the scenario MCP server. The plugins below register that server for the scenario skills; the standalone skill downloads can be used independently.
Option A: Plugin (recommended)
Claude Code
/plugin marketplace add integromat/make-skills
/plugin install make-skills@make-marketplace
Cursor — Team Marketplace: import integromat/make-skills and install make. The plugin registers the MCP server (/cursor).
Codex
codex plugin marketplace add integromat/make-skills
Then open the plugin directory, select the Make marketplace, and install make. If the MCP server is not registered automatically, add it manually:
codex mcp add make --url https://mcp.make.com/openai
codex mcp login make
Option B: Download the skills
Upload the zips to your agent (Claude Cowork, Claude Chat, and similar) or unzip them into its skills directory:
| Skill | Download |
|---|---|
| Bundle of all the skills | Download |
| Scenario Reference | Download |
| Scenario Explore | Download |
| Scenario Building | Download |
| Scenario Operations | Download |
| Make Endpoints | Download |
| Make Credential Requests | Download |
Skills directories, for agents that read skills from disk:
| Agent | Skills directory |
|---|---|
| ChatGPT / Codex | .codex/skills |
| Claude Code | .claude/skills/ |
| Cursor | .cursor/skills/ |
| Windsurf | .windsurf/skills/ |
| Cline | .cline/skills/ |
| Generic | .agents/skills/ |
Add the MCP server manually for scenario skills
Skip this if a plugin registered it. Each platform has its own endpoint, over HTTP:
| Platform | MCP URL |
|---|---|
| Claude Code | https://mcp.make.com/claude |
| Cursor | https://mcp.make.com/cursor |
| OpenAI Codex | https://mcp.make.com/openai |
claude mcp add --transport http make https://mcp.make.com/claude
claude mcp list
For any other agent, add the matching URL to its JSON MCP config.
On first use, authenticate through Make's OAuth consent screen. The server asks for one bundle of permissions and works only when every one of them is granted — if a tool answers with a permission error, reconnect and grant everything it asks for.
Troubleshooting
| Issue | Solution |
|---|---|
| MCP server not connecting | Check network connectivity to Make servers |
| MCP permission denied / 403 | Reconnect the Make MCP server and grant every requested permission |
| An MCP tool refuses a write | Read the refusal and the current tool contract; use the Make editor for capabilities the tool explicitly does not support |
| Endpoints SDK/CLI access failure | Check the selected key/zone, Make scopes, team access and provider connection separately; the MCP full-scope-bundle rule does not apply |
| Endpoints execution unavailable | Confirm organization beta/runtime access; local catalog discovery and package installation do not enable execution |
Contributing
Open pull requests against main — that's the trunk. Use squash merges and Conventional Commit PR titles (feat:, fix:, docs:, …), since release-please relies on them. A separate latest branch is fast-forwarded to each released tag. The Claude Code marketplace entry uses a local ./plugins/make-skills-claude source, so installs track the branch the marketplace is added from. Codex and npx skills add still resolve main HEAD directly. For Cursor Team Marketplace, import this repo and track branch main (branch is not set in .cursor-plugin/marketplace.json).
License
MIT
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.