Token-Based Credits for AI Actions

Every AI action you enable rides along with every message your AI sends — its tool definition is part of each request. Today we start counting those tokens toward message credits, fairly and transparently. And because heavy configurations now pay their own way, we're raising the Professional plan limit from 15 to 20 AI actions per agent.

Why We're Doing This

MCP tools are powerful because they carry rich schema definitions — but that makes them token-heavy, and the AI reads all of it on every request.

🔌 A lead form or button costs a few dozen tokens.

🔧 One complex MCP tool can cost 1,000–2,000 tokens.

🧩 Ten MCP-backed actions can add 11,000+ tokens to every message — several times the message itself.

That cost used to be invisible and unmetered. Metering it fairly is what lets us keep expanding AI actions instead of restricting them — so we can offer the full catalog of 834+ MCP apps and raise limits rather than tighten them.

How It's Charged

🎁 First 500 tokens free — most lightweight actions fit entirely within it.

📊 +0.5 content units per 2,500 tokens beyond the free allowance — so every 5,000 tokens equals 1 content unit, the same as user info and history.

✅ Only enabled actions count — disabled actions are never sent and never billed. Toggle one off and its tokens drop from every future message instantly.

At a glance: up to 500 tokens → free • 501–3,000 → +0.5 unit • 3,001–5,500 → +1.0 unit • 5,501–8,000 → +1.5 units, and so on.

You See Everything Before You Pay

🏷️ Per-action token badges — each action shows its exact token size, MCP actions included.

📈 Live cost meter — see your total against the free allowance and the next tier as you toggle actions.

🧮 Updated credit estimator — the per-message estimate now includes AI action units, line by line.

🔍 Debug breakdown — every message records the exact action tokens and units charged.

Professional Plan: Now 20 AI Actions per Agent

Fair metering means we no longer need to keep action counts tight. Professional workspaces can now enable 20 AI actions per agent, up from 15 — connect more MCP servers, API calls, and workflows to a single chatbot.

Where to Look

Open Settings → AI Actions to see each action's token cost, or Settings → Limits Control for the full per-message estimate. If a chatbot costs more than expected, check for MCP actions with large schemas — trimming unused tools is the fastest way to drop a tier.