Pipedream built its name on workflows, but those shut down on March 31, 2027; what it sells now is Connect, a hosted token vault, OAuth flow and MCP server in front of 3,000+ APIs. For one person the core is small: keep tokens for the handful of services you use, refresh them, run webhook and cron handlers, and expose a few tools to your agent. An agent builds that in a weekend. What it can't copy is the breadth: OAuth apps already registered for hundreds of providers, and 10,000 prebuilt tools that someone keeps current as those APIs change.
Before you buildFree is development only, so production traffic needs Startup. If all you need is webhooks and cron jobs for yourself, self-hosted n8n or Windmill covers that without a build.
Build me a personal integration gateway to replace Pipedream for my own scripts and agents. Requirements: - Node.js 22 + TypeScript: a Hono server on http://127.0.0.1:4880 with a small server-rendered admin page (vanilla JS) and a JSON API. `npm start` runs everything. - Storage: better-sqlite3 in ./data/gateway.db for connections, handlers and run logs. Access and refresh tokens are encrypted with AES-256-GCM using GATEWAY_KEY from .env. - Connections: OAuth 2.0 authorization code with PKCE for GitHub, Google (Gmail, Calendar, Sheets) and Slack, using client IDs and secrets I register myself and keep in .env; plain API-key connections for anything else. Refresh tokens 5 minutes before expiry and mark a connection "needs reauth" when a refresh fails. - Proxy: POST /proxy/<connection>/<path> forwards the request to the provider with a fresh token injected. A token never appears in a response, a log line or an error message. - Handlers: TypeScript files in ./handlers/, each exporting a trigger (webhook name or cron expression, via `croner`) and run(event, ctx); ctx.fetch(connection, url) goes through the proxy. Run each in a worker thread with a 30-second timeout and log input, output, duration and errors for 30 days. - Webhooks: POST /hooks/<name> with an HMAC signature check per handler (secret in .env), a 1 MB body limit, and a replay button for any logged event. - MCP: expose every handler and each connection's proxy as tools on a Streamable HTTP MCP endpoint at /mcp (`@modelcontextprotocol/sdk`), guarded by a bearer token from .env. - Single user. No accounts, no telemetry, local-first; outbound calls go only to providers I connect. - Out of scope: a visual workflow editor, per-user connections for other people, and the other 3,000 apps (a provider is added by editing a config file). - README: how to register each OAuth app (a Google consent screen left in Testing status expires refresh tokens after 7 days), exposing webhooks with a tunnel, and backing up the database and GATEWAY_KEY together.
Opens your agent with the prompt filled in, you press enter. Or copy it and paste it into an empty folder.
The first prompt builds the core. Run these one at a time in the same folder to grow it into your own Pipedream.
Add a Pipedream importer to the gateway in this folder. Requirements: - A command `gateway import-pipedream <zip>` that reads a project_export zip: one JSON file per workflow plus manifest.json. - For each workflow with status `exported`, write a TypeScript file in ./handlers/ with the trigger (HTTP, cron or an app event as a TODO), and one commented block per step holding the original step type, props and code. - Copy Node.js code steps as functions where possible; mark app steps that need a provider as TODO with the app name. - Print the manifest warnings and every skipped workflow in a summary; never overwrite a handler that already exists. - Update the README with how to export a project from Pipedream.
Add a small data store to the gateway in this folder. Requirements: - ctx.store.get, set, delete and list(prefix), backed by a SQLite table scoped per handler. - Optional TTL in seconds per key; expired keys are removed by a nightly job. - Values are JSON, at most 256 KB each; reject anything larger with a clear error. - An admin page that lists keys per handler and lets me edit or delete a value. - Update the README.
Add failure alerts to the gateway in this folder. Requirements: - When a handler run fails or times out three times in a row, send one message to an ntfy topic or a Slack incoming webhook from .env, with the handler name and the last error. - Send a recovery message when the next run succeeds; never send more than one alert per handler per hour. - Show the current failure streak per handler on the admin page. - Update the README.
Add Notion as a provider to the gateway in this folder. Requirements: - A config entry for Notion's OAuth 2.0 flow with NOTION_CLIENT_ID and NOTION_CLIENT_SECRET in .env, and the Notion-Version header sent on every proxied call. - Check Notion's docs for whether access tokens expire; if they don't, skip refresh for this provider and say so in a code comment. - A sample handler that appends a row to a database on a webhook call. - Update the README with how to create a public integration in Notion.
Teams building agents or products that act inside their users' accounts pay so they never register OAuth apps for dozens of providers, wait on app reviews, or keep token refresh alive. They also get a catalog of tools already wired to those APIs. For one person with five accounts that is a lot to buy; for a product with hundreds of users it is the shortcut.
Estimate · Runs on your machine or a $5 VPS. The build is token usage on your agent plan; each extra provider is a config entry plus an OAuth app.
| Plan | Monthly | Annual, per month | What you get |
|---|---|---|---|
| Free | $0/workspace | $0/workspace | Everything in development: 3,000+ connectors, 10,000+ prebuilt tools and triggers, MCP servers, the Connect proxy, custom OAuth clients and unlimited team members. No production environment; external users and usage credits count in development only. |
| Startup | $150/workspace | $99/workspace | Production environment, 100 external users included then $2 each, 10,000 usage credits a month then $0.012 each, email support. |
| Business | custom | — | Everything in Startup plus volume pricing for users and usage, SSO, SCIM, HIPAA, dedicated static egress IPs, SLAs and a dedicated Slack channel. |
Free has every feature, 3,000+ connectors and unlimited team members, but only in development: there is no production environment.
Kinda. The core is a weekend build with the prompt on this page, but real gaps remain: Hosted OAuth for 3,000+ APIs without registering an app at each provider, 10,000+ prebuilt tools and triggers that Pipedream keeps working.
About $150 a month on the Startup plan, $1,800 a year (checked 2026-10-19).
Hosted OAuth for 3,000+ APIs without registering an app at each provider; 10,000+ prebuilt tools and triggers that Pipedream keeps working; Managed runtime, scaling and a SOC 2 Type 2 attestation; Per-user connections: letting other people link their own accounts; A visual workflow builder, which is closing anyway. If any of those matter to you, keep paying.
Yes: n8n (free), Windmill (open source), Activepieces (open source).