See Alkimi at work, from source material to shared assistants.
Inspect a cited answer, revise a generated quiz, and see what students open when you share an assistant. Then explore the settings, models, and campus guardrails behind each workflow.
Create an account, then configure your assistant. Review plans. Departments and campuses can discuss institutional fit.
Designed for course and campus work across the sciences, humanities, and professional programs.


Build assistants from your course materials.
Ground assistants in selected course materials, set how they guide learning, and draft revisable assessment artifacts directly in chat.
Ground answers in your materials
Assistants answer from the sources you select, and show which source each answer came from.
Connect a collection, then decide how answers draw on it.
An assistant’s Knowledge page lists the collections it can search. For each one you choose whether answers may cite references, show figures, and offer downloads, and a recall control trades how widely each message searches against what it costs, with the estimate shown alongside.
One maintained collection can serve several assistants; update it once and every connected assistant answers from the new version.


-
Knowledge collections
Group course or campus content into collections that one or many assistants can draw on.
-
Documents, text, and websites
Upload PDF, images, Office and OpenDocument files, spreadsheets, LaTeX, Markdown, plain text and source code, paste raw text directly, or crawl a site with depth and prefix rules.
-
Cloud storage sync
Build and refresh collections from Google Drive, OneDrive, Dropbox, or Box.
-
Page-anchored blocks
Every document is read page by page into paragraphs, tables, figures and code blocks, so agents can show a figure inline and cite the exact block.
-
Retrieval control
Set how widely an assistant searches, keep a collection mention-only, and decide per collection whether it may cite passages, show figures, or offer files for download.
-
Citations in answers
Responses can cite the specific document, page, and passage they drew from.
-
Scoped questions
Students can aim a single question at one collection with an @ mention.
Cloud storage sync refreshes periodically rather than mirroring in real time. A citation makes a claim inspectable; it does not guarantee correctness.
Decide how the assistant supports learning
Set how your assistant guides learning, while administrators establish campus-wide policy guardrails.
Start from a teaching template, then make it yours.
The template catalog groups starting points by what they are for: Socratic discussion facilitators, exam prep and lab report coaches, citation helpers, and discipline-specific learning assistants. Each one is an editable configuration of persona, tone, rules, and tools, not a locked product.


-
Template catalog
Start from built-in education templates, including Socratic tutoring, course design, and assessment authoring.
-
Instructions Studio
Compose persona, tone, and reusable instruction sections with a live preview of the assembled prompt.
-
Safeguards
Attach prebuilt identity, data, content, and scope protections, with organization-wide defaults.
-
Reasoning visibility
Enable multi-step reasoning, and choose whether students see the thought process or only the answer.
-
Model choice
Give each assistant an Alkimi router that tracks the best model in its tier, or pin a specific model from the set your institution allows, and optionally let users switch mid-conversation.
-
Message spend limit
Set a model-message spend limit that also constrains available context. Enabled tools can add separate charges.
-
Web access control
Leave web search off, require an explicit @web request, or allow the assistant to search when it needs to.
-
Greeting and suggestions
Set the welcome message and the follow-up prompts students see in an empty conversation.
Templates and safeguards shape behavior; they are not guaranteed enforcement, so test in chat preview. The message spend limit bounds model-message cost and context; enabled tools can add separate charges, and retrieval scope is a separate setting.
Build and export coursework
Draft quizzes, exams, and rubrics in chat, then export them in LMS-ready, print, and document formats.
Draft a quiz from the readings and export it for your LMS.
Ask for quiz items from the connected materials; the assistant returns a card you can preview with the answer key and revise in chat. The download menu serializes the same quiz to the format your LMS imports.
Requires the quiz tool enabled on the assistant and a model with tool calling. Spot-check the first import.


-
Quiz generation
Enable the quiz tool and the assistant drafts quiz items as revisable cards inside the chat.
-
LMS export
Download finished quizzes as QTI 1.2 or 2.1, Moodle XML, or Blackboard packages for import.
-
Print and document exams
Export an exam for print, LaTeX, or Word when it is not going into an LMS.
-
Quiz import
Bring an existing QTI, Moodle, or Blackboard quiz into chat to revise or extend it.
-
Rubric generation
Draft grading rubrics as revisable cards and download them as Canvas rubric CSV.
QTI, Moodle XML, and Blackboard packages download from the quiz card for LMS import. Print, LaTeX, and Word exams are a separate document workflow, not an option in that export menu.
Share the assistant. Show what the learner opens.
Share assistants through full-page links and supported website or LMS embeds, with access configured for the intended audience. Separately, share a read-only conversation snapshot when permitted.
Put it where students already are
Publish assistants through full-page links, website and LMS embeds, guest access, and read-only chat snapshots.
What a visitor opens when you embed it.
The embed script places the assistant as a chat widget on a website or LMS page that allows iframe content; a direct link opens the same assistant full-page. Who can reach it is decided first, on the assistant’s permissions page: signed-out guests, workspace members by default, or named people with explicit roles.
Placement only, not page-aware help or LMS integration. Source-backed guest answers also need public access enabled on the attached collections.


-
Dedicated assistant page
Every assistant gets a full-page URL you can hand out directly.
-
Website embed
Drop an assistant onto an existing page as a floating widget script or an inline iframe, with styling options.
-
LMS embed
Use the generated iframe snippet to place an assistant inside your LMS as an external tool.
-
Guest access
Allow people without an account to use a shared or embedded assistant, billed to the organization.
-
External invitations
Send invitation links that give someone outside your organization access to one specific assistant.
-
Chat share links
Publish a read-only snapshot of a conversation, restricted to the organization or open externally, with optional redaction.
LMS embed is the generated iframe placed as an external tool, not native LTI 1.3 or gradebook sync. Guest access must be enabled on the assistant, and on the attached collections for source-backed guest answers. External invitees need an eligible personal subscription; chats on public-enabled assistants are organization-funded.
A purpose-built chat for coursework
The student-facing surface, built for coursework rather than general-purpose chat.
Share a conversation on purpose, with names removed.
Conversations are private to their owner. Sharing one is a deliberate act: choose organization-only or external visibility, scope the share to a point in the conversation, and let the dialog scrub the names it detected before a link exists.


-
Multi-assistant hub
One place to start and manage conversations across every assistant a student has access to, with search, filters, and bulk actions.
-
Temporary chats
Start a conversation that is not saved to your chat history, for quick questions that do not belong there.
-
File uploads in chat
Attach images and SVGs, PDFs, Office documents, spreadsheets, LaTeX, notes and source code inline in a message when the assistant is configured to accept them.
-
Math input and rendering
A LaTeX keyboard in the composer, and equations and proofs that render properly in the response.
-
Charts and plots
Interactive charts inline in the conversation, from function, parametric and polar plots to bar, distribution, 3D surface and network charts, with a builder for your own.
-
Durable generation
A response keeps generating if the student changes page or refreshes the browser.
-
Image generation and editing
Optionally allow inline image generation, billed and governed separately from text, with an in-chat editor to sketch, pin instructions, mask, crop and restyle images.
Temporary chats are not saved to account history; that is a client-level exclusion, not a zero-retention server guarantee. Image generation is an optional tool, billed and governed separately.
Govern it across the institution.
Set central guardrails, monitor usage and credit spend, and give every student accessibility settings they can turn on for themselves.
Set the guardrails once
Central controls that hold regardless of what any individual instructor configures.
One switch per model, for the whole campus.
Administrators decide which language models anyone in the organization can select. A model disabled here disappears from every assistant builder; a data-retention policy toggle sits above the catalog.
A configuration screen is a starting point for review. Check one consequence as a non-admin member.


-
Organizations and workspaces
Divide the institution into workspaces so departments, courses, and staff keep their assistants and materials separate.
-
Role-based permissions
Assign roles at the organization, workspace, assistant, and collection level.
-
Organization policies
Set rules for external sharing, public assistants, citation sharing, personal information, and default safeguards.
-
Model governance
Choose which models are available campus-wide, migrate assistants in bulk, and contact Alkimi to enable EU in-region routing for eligible models.
-
Shared templates and instruction library
Publish approved assistant templates and reusable instruction sections for the whole institution.
-
API keys and service accounts
Issue role-scoped keys so internal systems can work with Alkimi programmatically.
-
Billing controls
Manage the subscription, top up credits, review invoices, and set usage alerts.
Organization policies for sharing and access are enforced; safeguard defaults are starting points that local owners can adjust. EU in-region routing is requested through Alkimi for eligible models, not a self-service toggle.
See how it is being used
Clear visibility into active usage, credit consumption, and response quality across courses and departments.
Aggregate usage, by day, member, and assistant.
Organization reporting shows credit spend over time, broken down by category, with the top members and assistants ranked below. Export the tables as CSV for budget reviews.
Reporting never opens a private conversation.


-
Assistant statistics
Users, conversations, messages, response quality, performance, credit use, tool use, and a peak-hours heatmap.
-
Collection statistics
Content growth, usage over time, which passages get retrieved, storage, and the assistants drawing on the collection.
-
Organization usage
Credit spend broken down by day, member, assistant, and category, exportable as CSV.
Aggregate reporting does not open private student conversations, and response quality ratings do not measure learning gains.
Usable by every student
Every student can adjust interface scale, contrast, color vision support, and readable type from their own account.
Settings each learner controls for themselves.
Interface scale, high-contrast modes, color-vision compensation, and a dyslexia-friendly font live in account settings, so a student adjusts the interface without asking an administrator.


-
Interface scale
Scale the interface per device without relying on browser zoom.
-
High contrast
Raise contrast across the interface for low-vision readers.
-
Color vision support
Apply compensation filters for color vision deficiencies, or simulate them to check your own materials.
-
Dyslexia-friendly type
Switch the interface to the OpenDyslexic typeface.
These are per-account interface preferences that learners control themselves. They are not a Section 508 VPAT or a WCAG certification claim.
Start with one useful workflow.
Create a workspace and build your first grounded assistant, or talk with our team about rolling it out across a department or campus.
Create an account, verify your email, then configure your assistant. Review subscription tiers and credit allocations on the pricing page.