Tokens

Understanding how tokens work helps you use your AI budget wisely and avoid unexpected costs.

What is a token?

AI tokens are the unit of measurement used to track how much work your AI agents are doing inside Patchworks. Every time you interact with an AI feature - whether that's chatting with an agent, building a connector, or running a flow - tokens are being consumed.

---

How tokens are used in Patchworks

Token consumption varies significantly depending on what you're doing. Here's a practical breakdown, for one prompt against the following agents:

Agent	Typical token range
Flow Builder ( without Mapping )	36,000 - 40,000
Smart Mapping ( 1 Map request )	47,000 - 52,000
Connector Builder	90,000 - 120,000
Full Implementation ( Full Flow including multiple maps, connectors already exist )	92,000 - 110,000

---

Checking your balance and buying more

Every company has a monthly token allowance for AI Studio. You can see your current usage in the top-right of the Agents page. Token consumption varies by agent and task complexity - building a multi-step flow with mapping will use more tokens than a simple single-step flow.

Token limits are enforced per company. If you reach your monthly limit, AI Studio features will be temporarily unavailable until the next billing cycle or if you buy additional tokens. You can purchase more tokens from the button "Buy more tokens" on the main AI Studio screen.

Once payment is complete, your reference and top-up history are available in your Company profile under Allowances.

---

Practical guidance

Check your balance before large tasks

Before starting a connector build or a full implementation session, check your current token balance in Main side menu → AI Studio. If your balance is low, top up before starting rather than mid-task.

Break large tasks into stages

If you're building a complex connector with many endpoints, consider building it in stages. Completing the core authentication and a handful of critical endpoints in one session, then extending in a second, can make usage more predictable and easier to review.

Avoid redundant context

When chatting with agents, be specific. A focused question uses far fewer tokens than a broad one. Asking "why did run 4821 fail?" is more efficient than "can you look through my recent runs and tell me if anything looks wrong?"

Reuse what you've built

Connectors built with AI can be reused across flows and clients. The token cost is a one-time investment - once a connector exists, subsequent flows using it don't require rebuilding it from scratch.