MMdgtlQ: Expound on credits
Can you expound credit usage? I tried running an MCP and it's using 20+ credits each run somehow. Sometimes a single MCP instruction such as uploading on Wordpress have multiple runs and is using multiple credits too. Is this a bug? This will be a lot clearer if we can see how much credits an instruction will cost before do it AND a usage history.
Example: https://prnt.sc/IyPRqzE9JBLJ
https://prnt.sc/ca4k_FDRlFni
Last question is what's the ETA on MCP being on workflows? Thanks!
SeanP_AgenticFlowAI
Apr 25, 2025A: Hey there! Credits break down like this:
1. Token + compute fees
Every MCP call spins up an LLM to “think” about how to hit your WordPress (or any) API, so you pay for the actual tokens used plus a small flat compute charge. Complex tasks (e.g. parsing, error-handling loops) burn more tokens, hence 20+ credits per run isn’t a bug—it’s just proportional to the work.
2. Multiple steps = multiple runs
A single “upload to WordPress” action might actually:
• Fetch schema
• Format content
• POST the media
–…each sub-step costs credits.
3. Usage history & cost estimates
• Usage Dashboard: coming next sprint (you’ll see per-action breakdown).
• Pre-flight estimates: for Workflows we’ll offer draft credit forecasts—please vote for “Workflow Cost Estimates” on our roadmap so we’ll bump its priority.
4. MCP in Workflows ETA
We’re rolling out full MCP support inside the visual Workflow builder by early May.
In the meantime, if you’d like deep-dive help or to track our sprint progress, drop a vote here: https://agenticflow.featurebase.app/ and we’ll keep you posted!