API Workers

Overview

@cyanheads/mcp-ts-core/worker exports createWorkerHandler — the Workers entry point. It wraps tool/resource/prompt registries into a per-request McpServer factory that integrates with the Cloudflare Workers runtime.

createWorkerHandler(options)

import { createWorkerHandler } from '@cyanheads/mcp-ts-core/worker';
import { echoTool } from './mcp-server/tools/definitions/echo.tool.js';
import { echoResource } from './mcp-server/resources/definitions/echo.resource.js';
import { echoPrompt } from './mcp-server/prompts/definitions/echo.prompt.js';
import { initMyService } from './services/my-domain/my-service.js';

export default createWorkerHandler({
  tools: [echoTool],
  resources: [echoResource],
  prompts: [echoPrompt],
  setup(core) {
    initMyService(core.config, core.storage);
  },
  extraEnvBindings: [['MY_API_KEY', 'MY_API_KEY']],
  extraObjectBindings: [['MY_CUSTOM_KV', 'MY_CUSTOM_KV']],
  onScheduled: async (controller, env, ctx) => {
    // Cloudflare cron trigger handler
  },
});

Fresh scaffolds register definitions directly in the entry point as shown above. If your project later adds barrel files for definitions, importing arrays from those barrels is also fine.

Options

Key design points

• Per-request McpServer factory: a new server instance is created for each request. Required by SDK security advisory GHSA-345p-7cg4-v4c7.
• Env bindings refreshed per-request: Cloudflare may rotate binding object references between requests; the handler re-injects them on every call.
• OTel NodeSDK is disabled in Workers — canUseNodeSDK() returns false for V8 isolates, so no OTLP spans or metrics are emitted. Structured logs via ctx.log still work. OTEL_ENABLED=true has no effect in Workers. ctx.waitUntil() is received and passed through to app.fetch and onScheduled but not called by the framework (nothing to flush asynchronously).
• Singleton app promise with retry-on-failure: the framework init runs once; if it fails, the next request retries rather than leaving the Worker in a permanently broken state.

Binding types

Cloudflare Workers bindings come in two kinds with different injection mechanisms:

extraEnvBindings: array of [bindingKey, processEnvKey] tuples. The value of env[bindingKey] is assigned to process.env[processEnvKey] at request time.

extraObjectBindings: array of [bindingKey, globalKey] tuples. The object at env[bindingKey] is stored on globalThis[globalKey] at request time.

Both are refreshed on every request. Never cache binding references between requests.

CloudflareBindings extensibility

Core defines CloudflareBindings without an index signature, so servers extend it via intersection rather than module augmentation:

import type { CloudflareBindings as CoreBindings } from '@cyanheads/mcp-ts-core/worker';

interface MyBindings extends CoreBindings {
  MY_CUSTOM_KV: KVNamespace;
  MY_R2_BUCKET: R2Bucket;
}

Pass MyBindings as a type parameter where the framework accepts a generic env type (e.g., Hono route handlers, onScheduled).

Runtime compatibility

runtimeCaps feature detection

import { runtimeCaps } from '@cyanheads/mcp-ts-core/utils';

if (runtimeCaps.isWorkerLike) {
  // Workers-specific path
}

if (runtimeCaps.isNode) {
  // Node.js-specific path (e.g., filesystem access)
}

runtimeCaps is a snapshot taken at import time. Fields: isNode, isBun, isWorkerLike, isBrowserLike, hasProcess, hasBuffer, hasTextEncoder, hasPerformanceNow. All booleans, never throw.

Serverless storage whitelist

In Workers, only these storage providers are allowed:

filesystem, supabase, and unknown provider types are not on the whitelist:

• filesystem and unknown types throw ConfigurationError in serverless environments.
• supabase does not silently fall back. The serverless provider whitelist check fires immediately at the top of createStorageProvider() — Supabase credentials are never validated. Worker startup fails with ConfigurationError because Supabase is not on the serverless whitelist. Do not set STORAGE_PROVIDER_TYPE=supabase in a Worker.

Set STORAGE_PROVIDER_TYPE to one of the four whitelisted values to avoid unexpected behavior.

wrangler.toml requirements

compatibility_flags = ["nodejs_compat"]
compatibility_date = "2025-09-01"  # must be >= 2025-09-01

# Built-in storage providers require these exact binding names:
[[kv_namespaces]]
binding = "KV_NAMESPACE"       # required for cloudflare-kv storage
id = "..."

[[r2_buckets]]
binding = "R2_BUCKET"          # required for cloudflare-r2 storage
bucket_name = "..."

[[d1_databases]]
binding = "DB"                 # required for cloudflare-d1 storage
database_id = "..."

nodejs_compat is required for Node.js API shims (e.g., process.env, Buffer, crypto). The minimum compatibility_date activates the required shim set.

Binding names for core storage are hardcoded — the storage factory looks for KV_NAMESPACE, R2_BUCKET, and DB on globalThis. Using different binding names will cause a ConfigurationError. For custom (non-storage) bindings, use extraObjectBindings to map arbitrary binding names to globalThis keys.

Bundling requires a @duckdb/node-api alias. Wrangler's esbuild step follows the framework's lazy import('@duckdb/node-api') (the DataCanvas DuckDB provider) even though that path never executes on Workers, and fails on DuckDB's native platform bindings (Could not resolve "@duckdb/node-bindings-*/duckdb.node"). Alias the module to a local stub:

[alias]
"@duckdb/node-api" = "./duckdb-stub.ts"

// duckdb-stub.ts — rejects the lazy import if a misconfigured Worker ever reaches it
throw new Error('@duckdb/node-api is not available in Cloudflare Workers.');
export {};

Workers-specific warnings

instructions resolver runs after env injection. When instructions is a function, it runs inside initializeApp(env) — after injectEnvVars() — so env-derived text reaches the model without fighting the Workers module-load lifecycle:

export default createWorkerHandler({
  tools: [echoTool],
  instructions: (env) =>
    `Region: ${env.ENVIRONMENT ?? 'production'}.` +
    (env.MAINTENANCE_MODE ? ' Read-only mode — writes disabled.' : ''),
});

Plain strings work the same as on createApp. Type extends Omit<CreateAppOptions, 'instructions'>, so this is the only option whose shape differs between Node and Worker entry points.

Lazy env parsing is mandatory. Cloudflare injects env bindings at request time via injectEnvVars(), after all static module imports complete. Never parse process.env at module top-level in Workers:

// WRONG — parsed before env is injected
const apiKey = process.env.MY_API_KEY;  // undefined in Workers

// CORRECT — lazy parse inside a function or getter
export function getServerConfig() {
  return ServerConfigSchema.parse({ apiKey: process.env.MY_API_KEY });
}

in-memory storage is volatile. Data stored with the in-memory provider is lost between cold starts and is not shared across Worker instances. Use cloudflare-kv, cloudflare-r2, or cloudflare-d1 for any state that must persist or be shared.

Node-only utilities throw in Workers. scheduler (node-cron), sanitizePath (fs-based), and filesystem storage provider all throw ConfigurationError when called from a Worker. Guard with runtimeCaps.isNode or avoid entirely.

DataCanvas is unavailable in Workers. DuckDB has no V8-isolate build, so core.canvas is always undefined on Workers. Setting CANVAS_PROVIDER_TYPE=duckdb (the only non-default value) in wrangler.toml triggers a fail-closed ConfigurationError at init time:

DuckDB canvas requires Node.js or Bun. Set CANVAS_PROVIDER_TYPE=none or omit it for Cloudflare Workers deployment.

Leave the env unset (or set to none) for Worker deployments. Tools that conditionally use canvas should check the module-level accessor (if (!getCanvas()) { ... }) and surface a clear "feature unavailable on this deployment" message. See api-canvas for the full DataCanvas reference and setup wiring pattern.

The default notification bus is per-isolate. core.notify and the modern era's subscriptions/listen streams share a change-event bus that defaults to in-process — which on Workers means one bus per isolate, so a background emission reaches only the listeners that happen to live in the isolate that produced it. Handler-time ctx.notify* is unaffected (the listen stream and the handler are the same request). For fan-out that must cross isolates, supply a bus of your own, backed by a Durable Object or another shared channel:

createWorkerHandler({ tools, eventBus: myDurableObjectBackedBus });

The interface is the SDK's ServerEventBus — publish(event) and subscribe(listener). See api-context's notification section for what publishes onto it.

Testing Workers with miniflare

bun run test:worker runs the worker suite under tests/config/vitest.worker.ts (using @cloudflare/vitest-pool-workers + miniflare). Each test file gets its own fresh V8 isolate — module scope (including createWorkerHandler's appPromise singleton) is reset between files.

tests/config/vitest.worker.ts miniflare bindings

Declare all storage bindings used in the suite:

cloudflareTest({
  main: './tests/fixtures/worker-runtime.fixture.ts',
  miniflare: {
    bindings: { STORAGE_PROVIDER_TYPE: 'cloudflare-kv', ... },
    kvNamespaces: ['KV_NAMESPACE', 'CUSTOM_KV'],
    r2Buckets:    ['R2_BUCKET'],
    d1Databases:  ['DB'],
  },
})

Binding names must match the hardcoded names in storageFactory.ts (KV_NAMESPACE, R2_BUCKET, DB).

Per-provider test isolation

bindings.STORAGE_PROVIDER_TYPE is a global default. Per-provider test files override it by passing a modified env to worker.fetch():

import { env } from 'cloudflare:workers';

// Fresh isolate per file — appPromise starts null.
// First fetch initialises the singleton with the overridden provider type.
const r2Env = { ...env, STORAGE_PROVIDER_TYPE: 'cloudflare-r2' };
await worker.fetch(request, r2Env, ctx);

Use reset() from cloudflare:test in afterEach to clear binding state between tests. For D1, re-apply migrations immediately after each reset() (reset wipes the schema):

import { applyD1Migrations, reset } from 'cloudflare:test';

afterEach(async () => {
  await reset();
  await applyD1Migrations(env.DB, [{ name: '0001_schema', queries: [CREATE_TABLE_SQL] }]);
});

D1 schema setup

The cloudflare-d1 provider requires the kv_store table before any operations. Apply it in beforeAll via applyD1Migrations:

const KV_STORE_MIGRATION = `
CREATE TABLE IF NOT EXISTS kv_store (
  tenant_id TEXT NOT NULL,
  key       TEXT NOT NULL,
  value     TEXT NOT NULL,
  expires_at INTEGER,
  PRIMARY KEY (tenant_id, key)
)`;

beforeAll(async () => {
  await applyD1Migrations(env.DB, [
    { name: '0001_create_kv_store', queries: [KV_STORE_MIGRATION] },
  ]);
  sessionId = await openSession(d1Env);
});

R2 list limit

The R2 provider fetches limit + 1 objects to detect further pages. Miniflare caps R2 MaxKeys at 1000, so the default limit=1000 → request 1001 → error. Pass limit: 100 (or any value ≤ 999) to ctx.state.list() in test fixtures to stay under the cap:

const result = await ctx.state.list(prefix, { limit: 100 });

Storage probe pattern

Expose storage operations as MCP tools in the fixture to test them through the real HTTP surface:

const storageSetTool = tool('storage_set', {
  input: z.object({ key: z.string().describe('...'), value: z.string().describe('...') }),
  output: z.object({ ok: z.boolean().describe('Always true on success') }),
  async handler(input, ctx) {
    await ctx.state.set(input.key, input.value);
    return { ok: true };
  },
  format: (r) => [{ type: 'text', text: `ok=${r.ok}` }],
});

Call via tools/call over MCP JSON-RPC and assert on structuredContent. See tests/worker/storage-r2.worker.test.ts and tests/worker/storage-d1.worker.test.ts for the full pattern.