Intelligence · AI agents
AI automation tools that let you bring your own OpenAI or Anthropic API key
Zapier, Make, n8n, Pipedream and Activepieces all take an OpenAI or Anthropic API key of yours, and we found no markup on tokens you pay for that way. Each still counts the AI step on its own meter, and Make and Activepieces need a paid plan for your own key.
Zapier, Make, n8n, Pipedream and Activepieces all document a way to bring your own OpenAI or Anthropic API key into an automation, and we found no markup on the tokens you pay for that way on any of their pages. What you still pay each automation tool is its usual meter for the step that calls the AI: tasks in Zapier, credits in Make and Activepieces, executions in n8n, and compute time in Pipedream. Make's custom AI connections need a paid plan, Activepieces takes your own keys from its Plus plan up, and Zapier's AI by Zapier, Make's AI Provider and n8n's Gateway credits are no-key options that put the model cost on the platform's bill instead.
Everything this page says about those five products comes from their own pricing pages, help centers and documentation as we read them on November 6, 2026. We did not test the products or take screenshots, so open the linked page before you rely on a detail. We make Barside, a free desktop app for Windows, Linux and Mac that runs AI jobs on the key or plan you connect, and the later sections say what it does and where it does not fit.
Automation tools that take your own OpenAI or Anthropic API key, and what they still charge
AI automation tools that let you bring your own API key for OpenAI or Anthropic split the cost of one AI step in two: the AI company bills the tokens, and the automation tool bills its own unit for running the step. The table sums up what each vendor's pages say, and the sections below link the page behind every line.
| Tool | Your own OpenAI or Anthropic key | Plan for your own key | What the AI step still uses | The no-key option |
|---|---|---|---|---|
| Zapier | The ChatGPT (OpenAI) and Anthropic (Claude) apps, each with an API Key field | The two help pages name no plan rule | Tasks: 1 for a standard action step; we could not confirm the count for these app steps | AI by Zapier on Professional, Team and Enterprise, at 1x, 3x or 5x tasks by model tier |
| Make | Custom AI provider connections, such as OpenAI and Anthropic Claude | Paid plans | 1 credit per operation, while your provider bills the tokens | Make's AI Provider on all plans, with credits based on tokens and operations |
| n8n | A credential with your own API key on each AI node | No plan rule on the credential page; self-hosted n8n has no Gateway credits | 1 execution per run of the whole workflow | Gateway credits on n8n Cloud Starter and Pro, charged per request |
| Pipedream | OpenAI (ChatGPT) and Anthropic (Claude) apps that list an API key as the way to connect | The pricing pages name no plan rule | Credits by compute time, 1 per 30 seconds at 256 MB, with no charge per step | We found no AI credits of Pipedream's own |
| Activepieces | Your own AI provider keys | Plus and up on Activepieces Cloud | 1 credit per run, plus 1 credit per AI step with your own keys | AI steps at 2, 10 or 20 credits by model class |
Each tool does something well. Zapier's help pages walk you through both AI apps in seven short steps, and AI by Zapier needs no key at all. Make's credits page spells out which part of an AI run Make bills and which part your provider bills. n8n counts a whole run as one execution however many steps it has, Pipedream does not charge by step count, and Activepieces says in plain words that your own keys move the model cost to your provider's bill.
What BYOK means on an AI automation platform
A BYOK AI automation platform lets you bring your own key from an AI company, so the model use lands on your account with that company instead of inside the platform's units. One AI step then shows up on two bills: the AI company's, for tokens at its own rates, and the platform's, for the step in tasks, credits, executions or compute time. "No markup" only describes the first bill, because the platform does not bill those tokens.
The platforms' own AI options, in the last column of the table above, work the other way round, with the model cost folded into the platform's units. We found no published markup rate on the pages we read, and we did not open n8n's service pricing page, which holds its Gateway rates. Put the platform's price for a typical run next to your AI company's own price list before you decide which route costs less.
Your own key also brings a little work. You hold the account, add credit or a card with the AI company, watch its usage page and set its spending limits. In return, you can switch models or companies without touching the platform's plan, and one AI bill covers every tool that uses the key.
What the Zapier, Make, n8n, Pipedream and Activepieces docs say
The documentation of these automation platforms on bringing your own OpenAI or Anthropic API key sits on different kinds of pages, from help center articles and credential pages to app pages and pricing FAQs. Each vendor below gets the same three questions: how you connect a key, what the platform still counts, and what its no-key option costs.
Zapier
Zapier's help center has a getting-started page for ChatGPT (OpenAI) on Zapier and one for Anthropic (Claude) on Zapier. Both connect the same way: on the Apps page, choose + Add connection, pick the app, paste your key into the API Key field and continue. The OpenAI page asks for an OpenAI account with prepaid billing enabled, separate from ChatGPT plans, and says "you do not need a paid ChatGPT subscription." The Anthropic page asks for a paid Anthropic account with API access and a key made in the Anthropic Console. Neither app has triggers, so each one runs as an action, such as Send Message for Anthropic or Send Prompt for OpenAI.
Zapier's pricing page says triggers do not count as tasks, that its rate card shows 1 task for a standard step, and that AI and code steps vary by model tier, tool calls and runtime. The two app pages say nothing about tasks, so we could not confirm whether a ChatGPT (OpenAI) or Anthropic (Claude) action counts as a standard step. As of November 2026, the pricing page lists a Free plan with 100 tasks a month, and Professional from US$19.99 a month billed annually or US$29.99 billed monthly for 750 tasks, with prices set in US dollars.
The no-key route is AI by Zapier, which the pricing page describes as a way to "Pull the power of AI into your Zaps, without your own API key," on the Professional, Team and Enterprise plans. Zapier's article on AI by Zapier's model-based pricing sets the base cost per run by tier, "Standard 1x, Advanced 3x, Premium 5x," with tool calls added at the same rate, and new steps start on Advanced. The same article lists connecting your own AI account at 1x without naming providers, while the pricing page lists a bring-your-own-model (BYOM) tier and an Enterprise feature that routes "supported AI processing through your AWS Bedrock account." Ask Zapier which of these applies to your plan before you count on it.
Make
Make's credits page separates the two routes. With a custom AI provider connection, "You pay Make for credit usage based on operations, and pay your AI provider for token usage," and "Custom AI provider connections are available to users on paid plans." With Make's AI Provider, available on all plans in Make AI Agents and the Make AI Toolkit, the credits cover tokens as well as operations.
Make's page on credit usage for AI agents gives the same split for agents: "1 credit per operation + credits based on AI tokens" with Make's AI Provider, and "1 credit per operation" with a custom connection, where your provider bills the tokens. As of November 2026, Make's pricing page says the Free plan has 1,000 credits a month. The plan prices did not appear in the page text we read, so we could not confirm them.
n8n
n8n's Anthropic credentials page lists one authentication method, an API key. In the Anthropic Console, open Settings, then API Keys, create a key and paste it into the API Key field of the n8n credential; the Anthropic and Anthropic Chat Model nodes use it. Our Activepieces vs n8n self-hosting guide walks through n8n's OpenAI credential the same way.
On n8n Cloud Starter and Pro, Gateway credits let you run supported AI models without your own provider account, and n8n deducts each request from a prepaid balance at the rates on its service pricing page. You can switch any node to your own credential, and nodes on Gateway credits fail when the balance reaches zero. The page says Gateway credits aren't available on n8n Cloud Enterprise or self-hosted n8n, and it mentions no markup or fee.
n8n's pricing page explains its meter: "An execution is a single run of your entire workflow," and "It doesn't matter how many steps are in the workflow or how much data it processes." As of November 2026, Starter is 20€ a month and Pro 50€ a month, both billed annually, with 2.5K and 10K workflow executions. For n8n's Assistant, the pricing FAQ says "On self-hosted n8n, Assistant is available in preview. You bring your own key (BYOK), so no credit limits apply."
Pipedream
Pipedream's OpenAI (ChatGPT) page and Anthropic (Claude) page list an API key as the way to connect, with a Chat action on both. Both pages belong to Pipedream Connect, its toolkit for developers who add integrations to their own app or agent. Neither page says who bills the model use or mentions AI credits of Pipedream's own, and we found none on its pricing pages either.
Pipedream's Plans and Pricing docs also carry a notice: "Workflows and String are shutting down on March 31, 2027," and Pipedream Connect is not affected. Until then, workflows use one credit per 30 seconds of compute time at the default 256 MB of memory, and the page says Pipedream "does not charge for usage based on the number of steps." Development and test runs use no credits. The Pipedream pricing page now prices Connect: as of November 2026, Startup is $99 per month billed annually or $150 billed monthly, with 10,000 usage credits, and the page shows a dollar sign without a currency code.
Activepieces
Activepieces' Providers and keys page lists Anthropic and OpenAI among the providers you can connect with your own keys, and our Activepieces vs n8n guide covers the setup, so this part sticks to the bill. Activepieces pricing says "A run is one credit however many steps it has," and adds AI on top: 2 credits for a step on a fast model, 10 on a smart model and 20 on a frontier model, plus 1 credit for each agentic action. Its FAQ says "On Plus and up you can connect your own AI provider keys," which "drops every AI step to 1 credit" and moves the model cost "to your provider bill instead of ours." As of November 2026, Free is $0 with 1,000 credits a month and Plus is $20 a month, billed monthly, with 10,000 credits; the page shows a dollar sign without a currency code.
Which plan you need for your own OpenAI or Anthropic key
On an automation platform, bringing your own OpenAI API key or Anthropic API key (BYOK) depends on the plan in Make and Activepieces, while the n8n, Pipedream and Zapier pages we read set no plan rule for their key-based apps and credentials.
- Make: custom AI provider connections, such as OpenAI and Anthropic Claude, are for paid plans. On the Free plan, AI runs through Make's AI Provider, with credits based on tokens.
- Activepieces: your own keys start on the Plus plan of Activepieces Cloud. On Free, each AI step costs 2, 10 or 20 credits by model. Its Providers and keys page says your own AI keys "need a plan that includes AI providers," and we could not confirm whether the free self-hosted edition is one.
- n8n: each AI node takes a credential with your own key. Gateway credits exist only on n8n Cloud Starter and Pro, so on Enterprise and self-hosted n8n your own credential is the route the docs describe.
- Zapier: the ChatGPT (OpenAI) and Anthropic (Claude) help pages name no plan, and Free comes with 100 tasks a month as of November 2026. AI by Zapier, which needs no key, is on Professional, Team and Enterprise, and the AWS Bedrock route is an Enterprise feature.
- Pipedream: the pricing pages name no plan rule for API-key apps, and the docs say Free plans have a limited number of connected accounts.
An example month: the same steps, with the token bill on your own account
This is an example with assumed numbers, to show where the bill moves. Lena, a made-up operations lead at a small online shop, runs one support workflow 300 times a month. It has a trigger on a new support email, one AI step that asks Claude to sort and summarize the message, and two ordinary actions that add a row to a sheet and post a message to her team.
| Platform | With Lena's own Anthropic key | With the platform's own AI |
|---|---|---|
| Zapier | 900 tasks, if each of the three actions counts as a standard step (an assumption); Anthropic bills the tokens | AI by Zapier on the default Advanced tier: 3 tasks for the AI step, if 3x means 3 tasks with no tool calls (an assumption), plus 2 for the actions, so 1,500 tasks, with no key |
| Make, paid plan | 1,200 credits, if each of the four modules, trigger included, uses 1 credit (an assumption); Anthropic bills the tokens | The same 1,200 credits, plus credits based on the AI tokens |
| n8n | 300 executions; Anthropic bills the tokens | 300 executions, plus each AI request deducted from the Gateway balance on Cloud Starter or Pro |
| Pipedream | 300 credits, if each run ends within 30 seconds at 256 MB (an assumption), with the model use on her key | We found no Pipedream AI option on its pricing pages |
| Activepieces Cloud, Plus | 300 run credits plus 300 AI-step credits, so 600 | 300 run credits plus 3,000 for a smart model, so 3,300 |
Read across each row and the step count stays the same: the AI step is still a task, a module, part of an execution, compute time or a credit. What moves is the model cost, which leaves the platform's units and lands on Lena's Anthropic bill at Anthropic's own rates, which this example does not price. Compare the totals with your plan: 900 tasks is past the starting allowance of 750 on Zapier Professional, while 3,300 credits fits inside the 10,000 a month on Activepieces Plus and would be past the 1,000 on its Free plan.
Bring your own API key to AI agent workflow tools
Bring your own API key works in the AI agent features of these workflow tools too, with the same split between the platform's meter and your AI company's bill. The difference is that an agent decides how many steps to take, so the token bill grows with the agent's work.
- Make: running an agent with a custom connection uses 1 credit per operation, and your provider bills the tokens.
- n8n: AI nodes such as Anthropic Chat Model use your credential, and an execution still counts the whole run however many steps it takes.
- Activepieces: each agentic action is 1 credit on top of the run, and your own keys on Plus move the model cost to your provider.
- Zapier: its agents have a meter of their own, which our comparison of Make, Zapier and n8n AI features explains.
Set a spending limit on your AI company account before an agent's first scheduled run. A loop that takes more steps than you planned costs nothing extra on n8n's execution count, yet it still costs tokens.
How Barside works with your own AI key
Barside is a free desktop app for Windows, Linux and Mac, and it has no AI of its own. You connect Claude, ChatGPT or Gemini with an API key, or OpenRouter with a one-click sign-in, in Settings → AI, and that company bills you for what you use. If you already pay for a Claude or ChatGPT plan, My plan runs Claude Code or Codex on your computer, signed in with that plan. Barside's guides call an API key the safer choice, because Anthropic and OpenAI have not said that apps may use a plan this way.
Saved workflows and scheduled AI agents add no fee per run. Barside charges nothing for AI, and Barside Plus (US$20 a month or US$200 a year, plus sales tax) has no AI credits inside it. Your key is saved encrypted on your computer, and work on your computer sends requests straight to the AI company, with no Barside server in between.
- Download Barside for Windows, Linux or Mac and open Settings → AI.
- For Claude, ChatGPT or Gemini, Barside opens that company's key page in your browser. Make a key and copy it; Barside picks it up from the clipboard, checks it and saves it encrypted. For OpenRouter, sign in and press Authorize.
- Choose a model under the buttons. Our guide to getting a Claude API key for a desktop AI app covers the Anthropic side.
- Under AI spending limit, turn on Limit what Barside's AI spends and pick an amount from US$0.50 to US$10,000 for a day, a week (from Monday) or a month.
- Give the AI assistant a job, such as "Turn today's call transcript into a recap note with decisions and tasks" (an example request). Under its last reply, choose Save as workflow, then press Save workflow.
- To repeat it, open Agents & approvals → Workflows, choose Schedule, pick how often and a time, tick Ask me to run it on this schedule and press Save schedule. Each run waits in Agents & approvals → Inbox until you press Approve.
The spending limit estimates each answer from the AI company's list prices for your model, and OpenRouter reports the exact cost, which Barside uses. At 80% you get a notification and a banner, and at the limit the assistant, agents, Ask library, Quick AI and the Space helper stop sending requests until you raise it, turn it off or the period starts again. Each computer counts its own use.
With Barside Plus, cloud agents run your scheduled agents on Barside's server while your computers are off. Turning them on sends your key to that server, which checks it with your AI company and seals it with AES-256-GCM; the database keeps only the sealed bytes and the key's last 4 characters, and the key is used for nothing but your scheduled agents. Remove my AI key from Barside deletes it, and it is deleted by itself 30 days after Plus ends.
What Barside does not do
- No connectors between web apps. Barside has no app-to-app connectors or webhooks, so nothing starts when an email, a form entry or a new row arrives. A job like Lena's support workflow belongs in Zapier, Make, n8n, Pipedream or Activepieces.
- Your computer runs the work. Saved workflows run where Barside is open, and nothing runs while the computer is off, asleep or locked. Only the latest missed run from the last 12 hours is made up, and saved workflows stay on your computer even with cloud agents on.
- Cloud runs need an API key. My plan covers work on your computer, while cloud agents and web search inside Barside still need an API key. Cloud runs are not counted by the spending limit in Settings → AI, so set a limit on your AI account as well.
- Other apps. Typing and clicking in other apps works on Windows and Mac, and on Linux the assistant can focus windows and press keys but cannot read the controls inside other apps.
- Numbers. Up to 100 saved workflows with 40 recorded steps each, and 50 scheduled agents.
Which setup fits which job
These are made-up examples, to show how the facts above play out.
- Lena's support workflow starts in her email, so it needs an automation platform. On Make's paid plans, n8n or Activepieces Plus, her own Anthropic key keeps the model cost on her Anthropic bill while the platform counts the steps.
- Omar, a developer who already runs servers, self-hosts n8n. Gateway credits are not offered there, so every AI node uses a credential with his own key, and his executions do not grow with the number of steps.
- A team with flows on Pipedream workflows plans its move before March 31, 2027, and compares per-run meters while it rebuilds.
- Sofia, a consultant whose AI work is writing from her own notes, such as call recaps, morning briefings and drafts from her reading, runs it in Barside with her own key and a monthly spending limit. She keeps one automation plan for the web form that feeds her client list.
Our Zapier alternatives guide for solo freelancers counts three more workflows in Zapier tasks, Make modules and n8n executions.
Privacy: where your key and your data go
When you paste a key into an automation platform, that platform stores the connection and makes calls with it. Pipedream's pricing page says Connect credentials are "encrypted at rest, scoped per project," and "never exposed to your app." We did not read how Zapier, Make, n8n and Activepieces store connected keys, so ask each vendor before you connect a key with a large balance behind it, and give each tool its own key so you can revoke one without the others.
In Barside, your key stays encrypted on your computer and goes only to its own company, apart from the sealed copy cloud agents use. When a workflow or agent runs, the request and the notes the assistant reads go to the AI company you connected, whose data policy applies. Notes and workspaces you leave out of Library AI are never given to the assistant or its agents. Read the AI you connect and your data and how agents use your data, and our guide to AI apps where you bring your own key.
Works well with
Barside runs on Windows, Linux and Mac, and you can download it free.
FAQ
Questions and answers
Which automation tools let me use my own OpenAI or Anthropic API key?
Zapier, Make, n8n, Pipedream and Activepieces all document one. Zapier has ChatGPT (OpenAI) and Anthropic (Claude) apps with an API Key field, Make takes custom AI provider connections on paid plans, n8n uses a credential on each AI node, Pipedream's OpenAI and Anthropic app pages list an API key, and Activepieces Cloud takes your own AI provider keys from its Plus plan.
Do automation platforms add a markup when I bring my own API key?
We found no markup on tokens you pay for with your own key on any of the five vendors' pages. Make and Activepieces say your provider bills the model use. The platform still counts the step on its own meter: tasks in Zapier, credits in Make and Activepieces, executions in n8n and compute time in Pipedream.
Can I use my own OpenAI key in Zapier?
Yes. Zapier's help page for ChatGPT (OpenAI) asks for an OpenAI account with prepaid billing enabled and says you do not need a paid ChatGPT subscription. On the Apps page, add a connection, pick ChatGPT (OpenAI) and paste your key in the API Key field. Zapier's built-in AI by Zapier works without a key and is billed in tasks by model tier instead.
Can I use my own Anthropic key on Make's Free plan?
Make's help pages say custom AI provider connections, such as OpenAI and Anthropic Claude, are available on paid plans. On the Free plan, which has 1,000 credits a month as of November 2026, AI runs through Make's AI Provider, whose credits cover tokens as well as operations.
Does n8n charge extra when I use my own Anthropic key?
n8n's pricing page counts executions, one per run of the whole workflow however many steps it has, and a node with your own Anthropic credential adds no Gateway charge. Gateway credits are the no-key option on n8n Cloud Starter and Pro, deducted per request from a prepaid balance, and they are not offered on Enterprise or self-hosted n8n.
Is Pipedream shutting down its workflows?
Pipedream's Plans and Pricing docs say "Workflows and String are shutting down on March 31, 2027," and that Pipedream Connect is not affected. Until then, workflows use one credit per 30 seconds of compute time at the default 256 MB, with no charge by step count. If you build an AI workflow there now, plan where it moves next.
Can Barside use my own OpenAI or Anthropic key?
Yes. In Settings → AI you connect Claude, ChatGPT or Gemini with an API key, or OpenRouter with a one-click sign-in, and that company bills you. My plan runs Claude Code or Codex with a Claude or ChatGPT plan instead. Barside has no app-to-app connectors or webhooks, so it does not replace an automation platform for jobs that start in a web app.
Is Barside free?
Yes. Barside is free on Windows, Linux and Mac, and saved workflows and agents add no fee per run. The AI company you connect bills what you use, and an AI spending limit in Settings → AI caps each computer. Barside Plus is optional at US$20 a month or US$200 a year plus sales tax, and adds cloud agents, cloud sync and encrypted cloud backup.
Sources
Where the facts come from
Checked on November 14, 2026. Products change: if something here is out of date, tell us at support@barside.space.
- Zapier Help: How to get started with ChatGPT (OpenAI) on Zapier
- Zapier Help: How to get started with Anthropic (Claude) on Zapier
- Zapier Help: AI by Zapier new model-based pricing starting June 15, 2026
- Zapier pricing
- Make Help Center: Credits
- Make Help Center: Credit usage for AI agents
- Make pricing
- n8n Docs: Anthropic credentials
- n8n Docs: Use Gateway credits
- n8n pricing
- Pipedream: OpenAI (ChatGPT) API integrations and tools
- Pipedream: Anthropic (Claude) API integrations and tools
- Pipedream Docs: Plans and Pricing
- Pipedream pricing
- Activepieces pricing
- Providers and keys - Activepieces docs