Apps used to run your business.
Now your business builds the apps. 🧠
Your Second Brain, Fully Alive:
→ Projects remember
→ Agents reason
→ Workflows run
No code. No setup. Just creation.
→ taskade.com/ai
What is Workspace DNA?
Every Taskade workspace runs on one loop: projects remember, agents think, automations execute. Every app in the gallery runs on it too.
The AI steps in Taskade automations now run on your own OpenAI key.
Add the key in Connections > AI Model Keys.
Pick a TSK level in the AI step.
Taskade credits are not used.
Auto steps stay on credits. Pro and up.
taskade.com/automate
A freelancer's invoice generator, from one Taskade Genesis sentence. 💸
→ Clients and projects in one database
→ Invoices created and tracked to paid
→ Automations send the invoice and watch the payment
Day 3 of One Prompt, One App, 37 seconds:
youtu.be/cn-G65lG6uM
Your next CRM could be built by your sales manager.
Your next client portal by your account manager.
Your next dashboard by the ops team.
One prompt. One app. One system.
Powered by your Workspace DNA.
taskade.com/create
Replying to @marksmithzurich
Late follow-up on this. The Google Calendar steps in Taskade automations let you pick any of your calendars, your main one included, to list events, check free and busy time, or add new events. Questions anytime: [email protected]
You asked the right question in March: can it do the task as required. TSK-Bench is our answer. Every model built the same apps, and we used each app like a customer before we scored it. Every run: xitter.cf/Taskade/status/2111184…
🤖 Introducing TSK-Bench, our AI task leaderboard.
We tested top AIs on building apps, connecting agents, and running automations. Real tasks. Real scores. + Every run improves TSK-1.
🔑 Next up: bring your own AI keys to @Taskade!
🚀 Reply BYOK for early access!
Fair, the switch is the real cost. You do not have to move it all at once. A Taskade automation can watch ClickUp for new tasks or status changes and create matching tasks in a Taskade project. Both tools then run side by side while you test.
On token spend: Custom AI Agents can run on your own OpenAI, Anthropic or Google Vertex AI key on Pro and up. No Taskade credits on your own key.
Replying to @albertkinng @Samking207
Fair, and thanks for saying it plainly. Each page load starts on TSK-Auto, but your last pick is saved. Turn off the TSK-Auto switch in the model menu, and your model comes back.
For small edits, pick TSK-Instant. It uses the fewest credits of the four levels.
Replying to @Stylish_Lemur
Thanks for the BYOK reply. You are on the early access list.
BYOK works today in Custom AI Agents on your own OpenAI, Anthropic or Google Vertex AI key on Pro and up, and on your own xAI, Groq, Mistral or Together AI key on Business and up. In Automations, the AI steps run on your own OpenAI key, and the provider steps take your own OpenAI, Anthropic, Google Gemini, Groq or OpenRouter key. For Taskade EVE and app building it is coming soon.
To add a key, open Connections in your space, then AI Model Keys, and click Add key. What TSK-Bench found is in our newsletter: taskade.com/newsletters/w/L7…
Which provider do you want to use first? Questions anytime: [email protected]
Replying to @GeorgeWahbeh
Thanks, George. You are on the BYOK early access list.
Here is the exact status: BYOK works today in Custom AI Agents on your own OpenAI, Anthropic or Google Vertex AI key on Pro and up, and on your own xAI, Groq, Mistral or Together AI key on Business and up. In Automations, the AI steps run on your own OpenAI key, and the provider steps take your own OpenAI, Anthropic, Google Gemini, Groq or OpenRouter key. For Taskade EVE and app building it is coming soon.
No Taskade credits on your own key. The full TSK-Bench update is in our newsletter: taskade.com/newsletters/w/L7…
Which model do you want to bring?
Replying to @albertkinng
Good news: Flash is already on the board. The newest DeepSeek we tested is V4.1 Flash, and V4 Flash ran too. Open DeepSeek on taskade.com/tsk to see what each version built.
On OpenRouter: in Automations, the OpenRouter step already runs on your own OpenRouter key. Custom AI Agents take OpenAI, Anthropic or Google Vertex AI keys on Pro and up today.
No Taskade credits on your own key.
Taskade AI agents run on your OpenAI, Anthropic or Google Vertex AI key on Pro and up, and on xAI, Groq, Mistral or Together AI on Business and up.
Taskade EVE: coming soon.
taskade.com/agents
A correction to our reply above: Anthropic keys already work in Custom AI Agents. The full status: BYOK works today in Custom AI Agents on your own OpenAI, Anthropic or Google Vertex AI key on Pro and up, and on your own xAI, Groq, Mistral or Together AI key on Business and up. In Automations, the AI steps run on your own OpenAI key, and the provider steps take your own OpenAI, Anthropic, Google Gemini, Groq or OpenRouter key. For Taskade EVE and app building it is coming soon.
No Taskade credits on your own key.
A great prompt runs once. The system it creates runs forever. ⚡
→ The booking form still takes bookings at 2 a.m.
→ The overdue invoice still gets its reminder
→ The Monday digest lands, logged in or not
Taskade builds the system, not the answer.
taskade.com/automate
Your workspace can now go live. 🧠
Your second brain alive:
→ Workspace DNA
→ Persistent memory
→ Agents and automations
One prompt. One app. One living workspace.
→ taskade.com/ai
Replying to @albertkinng
Glad it lands, Albert. GPT leads TSK-Bench at 88, and DeepSeek is the only model rated Leading on interface. Which model do you want us to test next? Questions anytime: [email protected]
Late reply, and this one shipped. Taskade has a native MCP server now. Claude Desktop, Cursor and VS Code connect to your workspace through OAuth. Setup and the tools list: taskade.com/blog/taskade-mcp
Replying to @Samking207
Fair point, thanks for calling it out. Some of those models were out for weeks before they came to Taskade. We meant new in Taskade and wrote it badly. Not sure? TSK-1 Auto picks for you.

