TeachFX + Claude: Instructional Feedback Grounded in Real Classroom Talk

Almost every connector in the Claude for Teachers directory helps you produce something for students — a lesson, a worksheet, a diagnostic, a design. TeachFX is the one pointed the other direction. Its job in the Claude for Teachers connector directory is personalized instructional feedback grounded in real classroom talk — feedback for the teacher, about the teacher's own practice, sourced from what actually happened in the room. That's a quietly radical thing to put in a product, and it deserves its own piece.

Consider how thin the feedback loop on teaching normally is. A teacher might be formally observed a couple of times a year, by an administrator, for a single lesson, often a lesson everyone knows is being watched. From that, they're supposed to grow. Meanwhile the substance of teaching — the hundreds of small moves you make in classroom discourse every day, the questions you ask, the wait time you give, how much of the talking is yours versus theirs — goes almost entirely unobserved, because there's nobody in the room whose job is to watch it. TeachFX's contribution is feedback grounded in that real classroom talk. Not a checklist from a scheduled observation. The actual conversation of your actual class.

Feedback on teaching is the rarest resource in the building

Teachers give feedback constantly and receive it almost never. Students get comments on every assignment; the adult running the room gets a rubric a couple of times a year. This asymmetry is strange when you say it out loud, because the research on what improves outcomes keeps landing on things that live in the moment-to-moment texture of instruction — the quality of discourse, the balance of teacher and student talk, the kinds of questions that get asked. Those are exactly the things a twice-yearly observation can't see and a gradebook doesn't touch.

The reason the loop stayed thin was logistical, not philosophical. Nobody doubted that timely, specific feedback on your own teaching helps. It just required a second skilled adult observing your class often enough to matter, which no school can staff. TeachFX's connector matters because it grounds instructional feedback in real classroom talk without requiring that impossible second observer — the feedback comes from what your room actually did.

Why "grounded in real classroom talk" is the load-bearing phrase

Sit with that phrase, because it's doing all the work. Generic teaching advice is free and worthless — you can read a hundred blog posts about cold-calling and increasing student talk. What you can't get is advice about your Tuesday, in your room, with your students. Feedback grounded in real classroom talk is specific by construction. It isn't "teachers should ask more open questions" in the abstract; it's feedback anchored to the discourse that actually occurred. That specificity is the difference between advice you nod at and feedback you can act on the next morning.

This is also why it belongs in Claude for Teachers rather than as a standalone report. Feedback that reveals a pattern — say, that a particular part of your lesson consistently goes quiet — is most useful when the tool that surfaces the pattern sits next to the tool that helps you do something about it. In CfT, that next tool is right there: plan the adjustment, grounded in your standards and curriculum, in the same place you got the feedback.

Where the feedback becomes action

Here's how personalized instructional feedback grounded in real classroom talk turns into a changed lesson. This is the one bulleted list this piece will spend:

• See the pattern. TeachFX grounds the feedback in your real classroom talk, so what you're reacting to is your actual practice — not a generic best-practice checklist.
• Diagnose the moment. A pattern in discourse usually points at a design choice: a question that was too closed, a transition that killed momentum, a section where you did the thinking students should have done.
• Redesign with a partner. Bring that insight into Claude for Teachers and rework the lesson — grounded in the Learning Commons Knowledge Graph's standards and competencies — so the moment that fell flat gets rebuilt on purpose.
• Rehearse the moves. If the pattern is about questioning or dialogue, a discourse-focused skill like the Socratic tutor is a useful sparring partner for drafting the sharper questions you want to ask next time.

The point of the sequence is that feedback and improvement live in the same workspace. You're not reading a report in one app and planning in another. The observation about your practice and the redesign of your practice happen in one continuous move.

Teachers spend their careers giving feedback and almost never getting it. The rarest thing you can hand a teacher isn't another worksheet generator — it's an honest, specific mirror on their own instruction.

Feedback that respects the professional

Two guardrails make this a tool teachers can trust rather than fear. The first is stance. Personalized instructional feedback grounded in real classroom talk is coaching, not evaluation — it exists to help you get better, not to score you. That's consistent with the entire Claude for Teachers thesis: AI is here to strengthen the craft and give you back time for the human relationships at the heart of learning, not to sit in judgment. Use it the way a great instructional coach's notes work — as material for your own reflection.

The second is privacy, and it's non-trivial because classroom talk is sensitive data. Claude for Teachers runs on its own teacher terms built for K-12 privacy, protects student information under a K-12 Data Processing Addendum written to comply with FERPA, and does not use what you share for model training. You control what's shared. For a connector whose input is the literal conversation of your classroom, those aren't fine-print niceties — they're what makes it responsible to use at all.

The move that scales good practice

TeachFX connects to the deepest bet in Claude for Teachers: that the practices we know work — small-group instruction, differentiation, mastery pacing, rich classroom discourse — reliably improve outcomes and are expensive mostly in time and support, not in mystery. We already know rich discourse helps. What's missing is a feedback loop tight enough for a teacher to keep improving how they run it, without a coach permanently stationed in the room.

Ground the feedback in real classroom talk, put it next to the planning tools, and the improvement loop finally fits inside a normal week. That's the same move the whole connector directory makes — take a practice that only works with more time or more staff than schools have, and make it feasible for one teacher on a Tuesday. If you're building out the professional-growth side of your stack, the marketplace's education agent skills are a natural companion.

Want the ecosystem view? Start with the series hub, or browse more builder insights.

Part of the Claude for Teachers series. Related: Snorkl + Claude: Insights on Classes, Assignments, and Progress · Small-Group Instruction at Scale, with an AI Planning Partner. Browse AI tutoring skills or more builder insights.