You already use ChatGPT to plan your day. You ask Claude to read long PDFs. The same assistants can now read your training.
As of this week, Kenso is a connector for both. Link your account once, then ask either one anything: how your bench is trending, whether your HRV says today's a push day, what to swap out when an exercise feels stale. They'll pull live data from Kenso and answer with what you've actually done — not a guess.
Why we built it
A lot of the work in training happens between sessions: planning a block, second-guessing fatigue, deciding what's next. Most of that thinking already happens inside an AI chat for a lot of people. We didn't want to ask you to leave that flow.
So instead of building a chat surface to compete with ChatGPT, we made Kenso's data available to the ones you already use. Same idea as letting your bank statement land in your tax software — it goes where the work happens.
What you can ask
A few examples that actually return useful answers today:
- "Pull my recovery insights and tell me whether to push intensity today."
- "Summarize my last four weeks of training. What should I change?"
- "My HRV is in my low bucket — does that historically hurt my performance?"
- "How is my chest volume trending compared to last month?"
- "Build me a Push/Pull/Legs program for hypertrophy and save it as a draft."
- "Log a freestyle workout: 4×8 squats at 225, 3×10 RDL at 185."
Behind the scenes Kenso exposes 30+ read tools (workouts, sets, recovery, training load, signal insights, programs) and a handful of write tools (create program drafts, log freestyle workouts, stage imports). Your assistant decides which one to call based on what you ask.
Connect in Claude
- Open claude.ai (web or desktop).
- Settings → Connectors → Add, paste
https://api.kensoforge.com/mcp. - You'll be bounced to kensoforge.com to authorize. Pick the scopes you're comfortable granting.
- That's it. Start asking.
Claude is MCP-native, so the integration is rich and fluid — no tool-call ceremony, just questions and answers.
Connect in ChatGPT
ChatGPT's MCP connector support is currently behind a developer-mode flag — flip it on and the flow is nearly identical to Claude's.
- Open ChatGPT (Plus or Pro plan) and go to Settings → Connectors → Advanced, then enable Developer mode.
- Back in Settings → Connectors, click Add custom connector and paste
https://api.kensoforge.com/mcp. - You'll be bounced to kensoforge.com to authorize. Pick the scopes you're comfortable granting.
- Start asking.
Developer mode is OpenAI's current gate for third-party MCP servers. Once they roll connectors out broadly, the steps will shrink to two.
What it reads, what it doesn't
You grant scopes individually. The defaults read training history, programs, profile, recovery signals, and personalized insights. Two opt-in write scopes let your assistant create program drafts (always saved to your Drafts inbox — never auto-activated) and log freestyle workouts.
Neither ChatGPT nor Claude polls your account. They only read when a question triggers a tool call. And you can revoke any assistant at any time from Kenso → Settings → Connected Apps.
A note on Pro
The connectors are part of Kenso Pro. Pro is purchased on the App Store or kensoforge.com — not from inside ChatGPT or Claude. (Apple and Anthropic each have their own rules about subscriptions inside chat tools; this is the simpler path.)
Try it
If you're already a Pro user, head to Settings → Connected Apps in Kenso, or jump straight to the setup pages: Claude · ChatGPT.
If you're not Pro yet, start a free 30-day trial. The connectors come with it.
Training is full of context that's easy to forget and hard to keep straight. Letting the assistants you already use see your training data turns that context into something they can reason about with you — instead of the other way around.