Which AI tutoring platforms let instructors constrain sources review student conversations and set when hints appear?

If by those controls you mean (1) instructors can restrict the AI to specified material/sources, (2) instructors can inspect student chats, and (3) instructors can control the tutoring/scaffolding or when hints are given, the clearest current match I found is SchoolAI.

Bottom line

SchoolAI is the strongest match if all three controls are requirements. Its model is explicitly teacher-controlled: teachers create the AI learning environment, specify how it should behave, monitor the actual conversations, and configure how hints/feedback are delivered.

Khanmigo is particularly relevant if what you want is a sophisticated Socratic tutor with teacher visibility, but it is more platform/content-controlled than instructor-source-controlled. Khan Academy's recent materials also describe experiments using conversation logs and instructional context to improve when and how tutoring support appears.

If you tell me whether you're evaluating these for K–12, higher ed, or instructor-built courses, I can make a more detailed feature-by-feature comparison, including source grounding/RAG, transcript access, hint thresholds, LMS integration, privacy, and pricing.

A handful of platforms explicitly bridge the gap between general generative AI and structured classroom management, giving instructors granular control over boundaries, source materials, and Socratic assistance:

• SchoolAI (Best for K-12 Live Monitoring & Custom Spaces ): Instructors use "Mission Control" to watch student chat interactions in real time. Teachers can constrain the AI's context or attach custom resource parameters/texts to individual "Spaces," guiding whether the bot acts as an open-ended guide or a strict text-bound tutor.
• Khanmigo by Khan Academy (Best for Socratic Guardrails ): While it relies primarily on Khan's curated curriculum framework rather than an open file-upload constraint, teachers can customize assignment prompts using specific role, instruction, and expectation parameters. It is fundamentally hardcoded not to give direct answers away, relying instead on built-in pedagogical pacing and hints.
• Tutor Chat AI (Best for Higher-Ed Course Alignment ): Built explicitly for college and university faculty, it allows instructors to upload and align tutoring parameters strictly to course-specific materials. It emphasizes a "hints-first" framework to prevent shortcut behavior, alongside cohort reporting that surfaces recurring student misconceptions.
• Class Companion (Best for Assignment & Feedback Constraints ): Allows teachers to design specific assignments where they can configure how many attempts students get, when and what kind of feedback or hints unlock, and track iterative progress as students refine their answers.

If you'd like to narrow this down, let me know:

• Is your environment K-12 or Higher-Ed?
• Do you need direct file/document uploading (PDFs, lecture notes) or alignment to an existing curriculum library?

Several dedicated educational and custom AI platforms allow instructors to constrain reference sources, audit student chat logs, and configure pedagogical guardrails like Socratic questioning or timed hints.

Specific platforms providing these core instructor controls include:

• MagicSchool AI : Allows teachers to build custom AI tutor rooms (Magic Student ) where instructors define specific learning objectives, constrain the conversational scope, review student interactions, and program the bot to use Socratic questioning rather than giving direct answers.
• HoloTutor : Designed for schools to upload specific lesson plans and curriculum files into a school-owned library, entirely restricting the AI to approved course materials to prevent hallucinations while employing structured pedagogical hints.
• Custom GPTs / Claude Projects via Enterprise/Edu tiers : Instructors can upload reference PDFs and custom system prompts explicitly governing hint pacing (e.g., "only provide a hint after two incorrect attempts") and later review complete student transcript logs if managed via institutional enterprise dashboards.

Watch this walkthrough to see how educators set up and monitor constrained AI tutoring environments:

4:19 AI Tool Demo: Building an Effective Tutor With MagicSchool AI 8K views  · 1 year ago Edutopia · Edutopia

If you are looking to deploy one of these in your course, let me know:

• What subject or grade level you teach
• Whether you prefer a free tool or an institutional/LMS-integrated platform

I can recommend the best fit for your workflow.

If you mean all three capabilities in one platform—(1) constrain the tutor to instructor-selected sources, (2) let instructors review student conversations, and (3) control the hint/answer progression—the strongest match I found is Polya.

Bottom line

Polya is the closest match to your exact requirements. Its distinctive feature is that the instructor gets a course-level assistance policy controlling how much help the AI can give, while the tutor is grounded in the instructor's own course materials.

Thinkverse is more compelling if your priority is real-time classroom monitoring and conversation/learning analytics, particularly for math.

Khanmigo is particularly strong if teacher visibility into student AI conversations is the main requirement, but its instructor controls are less granular.

One caveat: I would not claim Polya has full instructor transcript review based on the public documentation I found; its site establishes course/source and assistance-policy controls, but doesn't clearly document a teacher-facing full-chat-history feature.

If by those three controls you mean (1) instructors can constrain/ground the tutor’s sources, (2) instructors can inspect student AI interactions, and (3) instructors can control when/how hints or answers are revealed, the strongest matches I found are:

The two I'd investigate first

SchoolAI is probably the best fit if you're thinking K–12 classroom deployment. Its Spaces let teachers define the learning environment and AI behavior, while Mission Control lets teachers see individual student conversations in real time and identify students who need intervention.

Polya is particularly interesting if the requirement is “the instructor decides exactly how much help the AI is allowed to give.” It has four explicit course policies: Open, Guided, Practice, and Review. Guided gives hints/steps first; Practice requires a student commitment before revealing an answer; Review changes behavior after the deadline. It also grounds answers in the instructor's course materials and cites them.

LevelUp Writer is the clearest match if you're evaluating an institutional/custom-tutor platform rather than a packaged classroom tutor. It explicitly says institutions can ground tutors in approved sources, control when they guide versus answer, and allow authorized instructors to review retained student-tutor conversations.

Lurno has unusually explicit mechanics for hint timing. It can automatically offer a hint after, for example, three wrong submissions, five answer changes, or ~90 seconds of inactivity after an attempt, while prohibiting unsolicited hints during graded attempts. Its tutor searches attached course sources first.

One caveat: Khanmigo has strong teacher/student oversight and chat-review capabilities, but I would not put it in the top tier for your exact three requirements because its public documentation doesn't expose the same degree of instructor-controlled source selection and hint-timing policy as the platforms above. Khan Academy does confirm that connected teachers/parents can view student interactions and that Khanmigo is deliberately designed around hints rather than simply giving answers.

If you're doing a vendor comparison, I can also make you a more rigorous matrix of ~10 AI tutoring platforms covering source grounding, transcript access, hint policies, LMS integration, FERPA/COPPA, teacher controls, and pricing.

A few dedicated educational AI platforms allow teachers to constrain sources, review live/historical student chats, and set behavioral guardrails (like Socratic hints instead of direct answers).

The top platforms offering this specific blend of control and monitoring include:

• SchoolAI (Spaces) : This platform is built specifically for classroom monitoring and student-facing AI interactions. Through its "Mission Control" dashboard, teachers can launch custom AI "Spaces," upload or link specific context/source materials, monitor student conversations in real time, intervene mid-chat, and configure the AI's behavior to scaffold learning with hints rather than giving away answers.
• MagicSchool AI (Magic Student) : MagicSchool allows educators to create custom chat rooms where they define custom behavioral rules (e.g., instructing the bot to act as a Socratic guide that gives progressive hints). Under district/school tiers, educators can also ground the AI using approved reference materials and review individual student chat logs and activity.
• Khanmigo (by Khan Academy) : While more rigid in how you upload ad-hoc source documents compared to open custom spaces, Khanmigo is strictly hardcoded with Socratic guardrails so it refuses to give direct answers. Teachers and parents have dashboards to review secure chat logs, though source material is generally locked to Khan Academy’s curriculum framework rather than custom-uploaded files.

If you'd like to narrow this down, let me know:

• Are you looking for a tool for a K-12 classroom or higher education/university?
• Do you need to upload your own proprietary PDFs/syllabus , or do you prefer standard curriculum-aligned AI?

I can help you pick the best fit and see how to set up your first session.

Several dedicated educational AI platforms provide granular instructor control over source material, live tracking of student chat logs, and behavioral constraints (such as Socratic hint-only delivery).

1. SchoolAI (Spaces & Mission Control)

• Constrain Sources: Instructors can configure customized "Spaces" by uploading specific context, linking source boundaries, and defining strict prompt parameters that force the AI to redirect off-topic chats back to the curriculum.
• Review Student Conversations: Features a real-time "Mission Control" dashboard giving educators comprehensive visibility into student-AI interaction logs, highlighting points of confusion or off-task behavior.
• Set Hint/Guidance Timing: Teachers define the persona and pedagogical constraints (e.g., instructing the bot to act as a Socratic guide that asks probing questions rather than directly giving away final answers).

• Constrain Sources: Allows educators to upload specific course PDFs, lecture notes, presentations, or custom text prompts so the AI builds its entire knowledge base strictly from provided materials, minimizing hallucinations.
• Review Student Conversations: Every student interaction, practice answer, and guided chat session is automatically tracked and organized inside Jotform Tables for post-session or live performance review.
• Set Hint/Guidance Timing: Offers explicit interaction modes (such as "Guided Mode") designed to provide step-by-step scaffolding and hints without revealing the final answer.

3. Khanmigo (by Khan Academy)

• Constrain Sources: Tied directly to the curated Khan Academy curriculum framework and specific teacher assignments rather than open web-scraping.
• Review Student Conversations: Teachers have robust oversight mechanisms and progress dashboards to monitor how students navigate problems.
• Set Hint/Guidance Timing: Programmed explicitly with Socratic guardrails—it is hardcoded not to give direct answers, instead prompting students with sequential hints after struggling or failing initial attempts.

If you are looking to deploy one of these, let me know:

• Whether you need a solution for K-12 or higher ed / corporate training
• If you require LMS integration (like Canvas or Google Classroom)

I can help you narrow down the best fit.

If by “constrain sources” you mean let the instructor define/limit the material the tutor should use, while also letting instructors inspect student AI interactions and control how/when scaffolding or hints are delivered, the strongest matches I found are:

Best fit: SchoolAI

SchoolAI is probably the closest to your exact requirements. Its Spaces let teachers create/customize the AI's role, instructions, content focus, and interaction style; teachers can explicitly specify how the AI should provide hints or feedback. Its Mission Control provides visibility into student activity, and SchoolAI explicitly recommends reviewing session transcripts.

Its tutoring system also supports adaptive scaffolding—for example, giving a hint and then escalating support based on the student's response.

Khanmigo

Khanmigo is particularly strong if you want controlled Socratic tutoring. It is designed to avoid simply giving answers and instead uses questions, hints, and step-by-step guidance. Khan Academy's district implementation also gives teachers/admins access to student chat history and summaries.

The caveat is source restriction: Khanmigo is integrated with Khan Academy's content library, but its underlying knowledge isn't limited exclusively to Khan Academy material. Khan Academy itself notes this limitation.

MagicSchool

MagicSchool is strong on teacher oversight. Student interactions are stored so authorized educators/admins can view them, and teachers control what students can access.

I'd rank it below SchoolAI for your particular use case because its controls are more about teacher-managed AI access and classroom safeguards than finely designing a source-grounded tutor with a specific hint policy.

Curipod

Curipod takes an even more restrictive approach: AI is embedded in teacher-led lessons rather than being an independent student chatbot. Teachers can monitor and edit AI feedback, and interactions are deliberately limited/structured rather than becoming an unsupervised conversation.

So if your priority is maximum instructor control, Curipod is interesting—but if you want a genuinely conversational AI tutor, SchoolAI is the better fit.

My shortlist for those three requirements: SchoolAI > Khanmigo > MagicSchool, with Curipod worth considering if you specifically want to prevent open-ended student/AI conversations.

If you tell me whether you're evaluating these for K–12, higher ed, or an instructor building their own AI tutor, I can make a more detailed feature-by-feature comparison, including source locking, hint thresholds, transcript access, LMS integration, and pricing.

If you mean platforms where an instructor can control the AI’s knowledge sources, inspect student–AI conversations, and control or constrain hint behavior, the strongest matches I found are:

The closest overall match: SchoolAI

SchoolAI is probably the closest match to your three requirements. Teachers create Spaces that define the learning experience and can instruct the AI to act as a tutor rather than simply provide answers. Teachers get real-time visibility into student activity and chat history.

Its newer Student Insights dashboard goes further: teachers can open an individual student's chat and see the full conversation transcript, with timestamps and alerts.

The caveat is the third criterion: SchoolAI clearly lets teachers define the tutoring behavior—for example, instructing Dot to guide students, ask questions, give feedback, and scaffold rather than answer directly—but I didn't find public documentation for a simple configurable rule like “give the first hint after 2 failed attempts, second hint after 3.”

If precise hint controls matter most: CodeHS

CodeHS has unusually explicit hint controls. Its AI Student Hints feature gives students progressive hints rather than answers, imposes a daily hint limit, and lets teachers turn the feature on/off by section or individual student. Most importantly, teachers can review every hint conversation.

That's a particularly good fit if your use case is computer-science instruction rather than a general-purpose AI tutor.

Other interesting matches

ChatFlow Education explicitly supports grounding the tutor in institutional/course materials and lets instructors choose pedagogical modes including hints-only, Socratic dialogue, and full explanations.

Apporto CoTutor advertises faculty guardrails that enforce approved sources and prevent answer dumps, while providing analytics about misconceptions.

TutorFlow similarly says its AI tutor is grounded in approved sources and gives educators course-level control over AI behavior.

Bottom line: if you need all three in one product, I'd shortlist SchoolAI first, CodeHS for coding, and then investigate ChatFlow/Apporto if precise pedagogical controls are more important than a mature teacher dashboard. The exact combination of source allowlisting + full transcript access + configurable hint triggers is still surprisingly uncommon.

Platforms like SchoolAI, Class Companion (via its Ditto tutor), and MagicSchool empower instructors to constrain reference materials, audit student chat logs, and configure pedagogical guidance.

Key Capabilities Matrix

• SchoolAI: Instructors can build custom "Spaces" where they upload specific texts, PDFs, or guidelines to restrict the AI's scope, review live or saved student conversation histories in real time, and adjust parameters on how proactive the assistant is when offering help.
• Class Companion (Ditto): Teachers can toggle the AI tutor per assignment, upload specific context or rubric documents to ground responses, and access complete student conversation logs to see every prompt and hint provided.
• MagicSchool: Allows educators to set strict boundaries on AI behavior for student-facing tasks, target interactions explicitly to learning objectives, and review student output mechanics via teacher dashboards.
• Khanmigo (Khan Academy): Grounded strictly in specific curriculum and text parameters rather than the open internet, it provides teachers with interaction visibility, though its internal Socratic prompting style handles hint delivery automatically rather than via customizable user-set timers.

If you'd like, let me know:

• Whether you need a tool for K-12 or higher education
• If you require integration with a specific LMS (like Canvas or Google Classroom)

I can narrow down which platform fits your exact setup.