toolscaled.

Works in Claude, ChatGPT, Gemini, Grok and any MCP agent

Do it once in chat. Run it whenever you need it.

You already get real work done in Claude, ChatGPT, Gemini or Grok. Toolscaled turns the conversation that worked into a workflow that runs on a schedule or when new work arrives, without the agent starting from scratch every time.

Sign up free Sign in

Free plan. No card. · See pricing · Read the guide

Works where you already work.

Claude·ChatGPT·Gemini·Grok

Add Toolscaled by its address. Sign in once. Your workflows are there in every chat.

Faster

Most steps don't need a model. Independent steps run at the same time.

Cheaper

You only pay for a model on the steps that need one, and those use a smaller, faster model.

Same every time

The steps stay fixed until you deliberately change them.

Easy to fix

Open any run in chat. Your agent sees exactly what happened and helps you fix it.

Your day today. Your day with Toolscaled.

  1. Today

    You open your AI chat. You paste the same instructions. You check the output. You fix the same mistake as last week. Tomorrow you do it all again.

  2. Day 1

    Work the way you already do. Solve the problem in chat, with your own tools. When the result is right, tell your agent: "Save this as a workflow."

  3. Setup

    No new app to learn. No boxes to draw. Toolscaled saved every step while you worked, and your agent turns those steps into a workflow.

    You decide which steps run on their own, which ones the model handles, and where you approve. Test it on the work you just did. Nothing runs until you turn it on. Any step can wait for your OK before it sends.

  4. One month later

    The work happens without you.

    The morning report is in your inbox at 7.
    A new form entry is logged and answered.
    A support email gets a reply that cites the order and the policy. The unclear ones wait for you.

    You stopped re-prompting. You stopped watching over it.

Your agent solves it once. Toolscaled runs it every day.

Three jobs people stop doing by hand.

Do the work in chat with the agent you already use, capture the procedure, then run it on fresh data. Keep a review step wherever you need one. Load any run into your agent to improve it.

chat — just describe it Each morning, email me a report on my watchlist On it — pulling quotes and last night's filings ⋯ Done ✓ — your report is in your inbox. Perfect — do that every morning at 7. ✓ captured as a workflow ⏰ 07:00 · daily get_watchlist() get_quote({ticker}) for each — ×3 tickers summarize_filings() per ticker · ×3 send_email(report) 📩 report delivered ✓ same process — fresh data 01 · CHAT — describe the problem. The agent does the work. 02 · CAPTURE — the agent wired it up. Here's the workflow it built. 03 · RUN — 07:00 fires, the graph replays. No agent in the loop. Every run: the same approved process, applied to fresh data chat — just describe it When my contact form gets an entry, log it in my sheet and draft a reply On it — trying it on the last entry, from Dana ⋯ Done ✓ — row added, reply sent after your OK. Great — handle every new entry like that. ✓ captured as a workflow 📝 contact form · new entry fetch_page({website}) append_row({entry}) Google Sheets · lead log draft_reply({entry}) model step · your tone send_email(reply) Gmail · after your approval 📧 row logged · reply sent ✓ every entry, handled the same way 01 · CHAT — describe the problem. The agent does the work. 02 · CAPTURE — the agent wired it up. Here's the workflow it built. 03 · RUN — a new form entry arrives. The same playbook replays. Every entry: logged, answered, and approved by you chat — just describe it Can you answer support emails? Check the order, our docs, then reply. On it — trying the latest email, from Dana ⋯ Done ✓ — drafted from the order + policy, replied. Great — handle every email like that — flag any you're unsure of. ✓ captured — unsure ones wait for your approval. 📧 new email · support@ lookup_customer({sender}) classify_issue({body}) ✦ get_orders({customer}) kb_search({issue}) draft_reply(order · kb) cites the order & the policy send_reply({thread}) ⏸ human review confident ✓ unsure 📧 replied in the customer's thread ✓ ✓ edge case — approved by a person, then sent 01 · CHAT — describe the problem. The agent does the work. 02 · CAPTURE — the agent wired it up. Here's the workflow it built. 03 · RUN — an email arrives. The same playbook replays. When it's unsure, it pauses for a person — approve, and it sends.

The systems in these demos are your own. Anything with an API or an MCP server connects.

how it works

Do it. Save it. Run it. Fix it.

01

Do the work in the agent you already use.

Connect your tools, and work through one real task in Claude, ChatGPT, Gemini, Grok or any MCP agent. Correct it until it's right. Toolscaled records every tool call, and a console inside Claude and ChatGPT shows you each step.

02

Check it, then publish.

Read what each step does and where it stops for you. Test it on the work you recorded: steps that send or write reuse their recorded results until you say go. Nothing runs on its own until you turn it on.

03

It runs when the work arrives.

On a schedule, a new email, a form entry, a webhook, after another workflow, or when you click Run. Fixed steps run on their own. Approvals wait for you.

04

When something breaks, open it in chat.

One list shows failed runs and approvals that wait for you. Load a run into your agent, and it sees exactly what happened. Fix the workflow in the same conversation and publish it. The next run uses the fix.

Shared workspaces with roles. Scoped API keys. Run history. A new version each time you publish.

how is this different

If your agent can do it, you can automate it.

vs. Zapier, Make, n8n

You don't draw boxes. You do the work once, in chat.

vs. scripts and cron jobs

No code to write or maintain. The conversation becomes the steps.

vs. a saved prompt

A saved prompt makes it up again every time. A workflow runs the same steps every time, and it uses a model only where you said it should.

what connects

Bring the tools you already use.

Google is built in: Gmail, Sheets, Calendar, Docs, Drive and more. Sign in once. Outlook is built in as an email trigger. HubSpot, Linear, Airtable, Calendly, Typeform and more connect in a few clicks. Anything with an API or an MCP server connects too: paste its address into chat, pick what you want, and sign in on Toolscaled. Your passwords and keys never go through chat.

01

Give your agent the address.

Paste an OpenAPI spec URL or document, or an MCP server URL, into Claude, ChatGPT, or your agent. In the app, you can also upload a spec file.

02

Preview what it can do.

Toolscaled reads the spec or tests the MCP server and lists its operations. Choose the ones you want. A preview creates nothing.

03

Sign in on Toolscaled.

Add an API key or token, or sign in with OAuth where the provider supports it. You enter credentials on the Toolscaled website, so secrets never pass through chat.

04

Use the new tools.

Each operation becomes a named tool, for example helpdesk-list-tickets. Your agent works with it, and your workflows repeat it.

what starts a run Run on a schedule, when a new email or form entry arrives, when a signed webhook fires, when another workflow finishes, or when you start it yourself.

Pick one task you do every week. Make it a workflow.

Start with something you already do in Claude, ChatGPT, Gemini, Grok or your own agent. The Free plan needs no card.

Built for teams that already use Claude and ChatGPT.

Sign up free Sign in

See plans and pricing · Read the guide

Want help building your first workflow? Talk to us.