make-testkit

Test Make.com (Integromat) scenarios offline — no account, no network, no waiting on a real run to find out a mapper was wrong.

Make scenarios are real software: filters, routers, iterators, aggregators, IML formulas, and error-handling routes. But they live in a web UI with no unit tests, so a dropped guard or a typo in a {{...}} mapper ships silently and you find out when leads get double-pushed or an email goes out blank.

make-testkit runs your exported blueprint JSON through an engine that reproduces Make's execution semantics — including failure semantics — so you can assert behavior in CI the same way you would any other code.

┌─────────────────┐   ┌──────────────────┐   ┌────────────────────┐
│ @make-testkit/   │   │ @make-testkit/    │   │ Claude Code plugin  │
│ engine           │──▶│ mcp (MCP server)  │──▶│ (slash commands)    │
│ run · eval · lint│   │ 4 tools over stdio│   │ /make-iml /make-… │
└─────────────────┘   └──────────────────┘   └────────────────────┘

What's in the box

• @make-testkit/engine — the core. An IML expression evaluator, Make filter operators, an Airtable filterByFormula evaluator, a blueprint interpreter that models routers / aggregators / iterators and the five error directives (Resume, Ignore, Break, Commit, Rollback) plus consecutive-error auto-disable, and a static linter + contract checker.
• @make-testkit/mcp — a Model Context Protocol server exposing four tools: eval_iml, simulate_blueprint, lint_blueprint, run_testsuite.
• Claude Code plugin — /make-iml, /make-simulate, /make-lint, /make-test, wired to the MCP server.

Quickstart

git clone https://github.com/SaharBarak/make-testkit
cd make-testkit
npm install
npm run build   # builds the MCP server
npm test        # runs the engine + example suites

Use the engine

import { runBlueprint, lintBlueprint, checkContracts } from "@make-testkit/engine";
import { makeState, handlers, contracts } from "./examples/handlers.mjs";
import bp from "./examples/blueprints/lead-sync.json" with { type: "json" };

// 1. simulate a run
const ctx = { state: makeState(), handlers };
const res = await runBlueprint(bp, ctx);
console.log(res.status);          // "success" | "warning" | "error"
console.log(ctx.state.pushed);    // what the scenario actually did

// 2. lint for structural defects
console.log(lintBlueprint(bp));   // unresolved refs, duplicate ids, …

// 3. pin behavior with config-drift contracts
console.log(checkContracts(bp, contracts).ok);

Your service modules (Airtable, an HTTP API, an LLM, …) are plugged in as handlers keyed by Make module name, or stubbed per test so every external assumption stays explicit. Builtins (router, aggregator, iterator, set-variables, JSON, webhooks) work out of the box.

Use it from Claude Code

/plugin marketplace add SaharBarak/make-testkit
/plugin install make-testkit@make-testkit

Then, in any project that has Make blueprints:

/make-lint ./scenarios/lead-sync.json ./scenarios/contracts.mjs
/make-simulate ./scenarios/lead-sync.json ./scenarios/handlers.mjs
/make-iml {{lower(trim(1.Email))}}

Use the MCP server directly

Add to any MCP client (e.g. claude mcp add):

{
  "mcpServers": {
    "make-testkit": { "command": "node", "args": ["/abs/path/make-testkit/packages/mcp/dist/index.js"] }
  }
}

How it models Make

The interpreter follows the documented Make execution model: bundles process serially depth-first; a module error with no onerror route stops the run (rollback) — unless the scenario has DLQ enabled and the error is mid-run, in which case it stores an incomplete execution and the run is a warning; error routes expose {{failedId.error.message}} and resolve to whichever directive they reach; aggregators emit a single { array, __IMTAGGLENGTH__ } bundle even when empty. See packages/engine/src/interpreter.mjs for the full notes.

It is a faithful subset — it covers the module types and IML functions real scenarios use, and fails loudly on anything it doesn't model rather than guessing. Extend it with registerFunction, registerOperator, and your own handlers.

Docs

Full guide and API: https://saharbarak.github.io/make-testkit/

License

MIT © Sahar Barak