Talk to your training data from any AI tool
Connect Claude Desktop, Claude Code, Cursor, or any MCP-aware AI agent to your Konstantly workspace. Ask compliance questions, automate assignments, and analyze training data, with the same permissions as the browser UI.
Claude Desktop
macOS, Windows
Claude Code
CLI
Cursor
AI editor
ChatGPT
Rolling out
Slack bots
Internal agents
Custom agents
Any MCP client
Audit prep in a sentence
Stop reverse-engineering questions through filters and CSV exports. Ask the agent what you actually want to know, and your permission set decides what comes back.
- OSHA / HIPAA reporting · "Who is overdue on confined-spaces training across the warehouse group?"
- Certification expiry · "What is expiring in the next 60 days? Schedule recerts two weeks before."
- Weekly digests · "What training was completed last week? Format as a 4-bullet exec summary."
- Audit-log queries · "Show me every assignment created by the agent last quarter."
Ask Claude anything…
Onboarding without the click-through
Bulk-assign training, offboard departing users, change email addresses, from inside the AI tool your team already uses. Every action is audited identically to the UI.
- One-prompt assignments · Single course to one learner, or a learning path to a whole group.
- Learning paths · Push full sequences with ownership checks already in the platform.
- Offboarding hygiene · Ban and unban users with role hierarchy enforced.
- Account maintenance · Update an email address, and the platform handles the notification.
Ask Claude anything…
Drop a PDF. Get a course.
Hand the agent your source material (a vendor PDF, a policy doc, a set of slides) and ask for a course. It builds the structure, writes the lessons, and adds a final quiz, grounded in what you uploaded. Available on Business and Enterprise.
- Upload, then generate · PDF, DOCX, or image. The agent extracts text and uses it as the basis for the course.
- Or just describe it · Skip the upload and give a prompt, and it works the same way, with the same audited result.
- Structured output · Title, modules, lessons, pages, final quiz. Same shape as a hand-built course.
- Author stays in control · Open the course in Konstantly to review or edit before publishing. You're the editor.
Ask Claude anything…
Tell it what to change. It changes it.
Generated a course but it's missing a section? Question landed flat? The agent can extend, rewrite, and refresh, without you opening the editor. Append-only by design, so existing content is never lost.
- Extend, don't replace · Add new lessons, tests, or sections. The agent appends; existing content stays intact.
- Fix one question · Wrong answer? Too easy? Distractors too obvious? Rewrite a single question on the same concept.
- Refresh a lesson · Update a lesson's brief (topic, key points, audience) without touching pages you've already authored.
- Always audited · Every AI-driven change writes a dedicated audit event. Previous content stays in the audit log for review.
Ask Claude anything…
Same posture as the browser UI
MCP doesn't introduce a parallel permission system. It inherits everything from your existing roles. Revoke a permission and the next MCP call respects it immediately, with no cache to invalidate.
- Per-user keys · Each person issues their own from Settings → MCP Keys.
- Scoped + license-aware · Each key carries read or read+write scope. License clamps apply dynamically, no re-issue required.
- Revoke in seconds · Lost laptop? Revoke from the dashboard; next call from that device fails auth.
- SIEM-ready audit trail · Every key event and write action lands as a dedicated, queryable event.
Ask Claude anything…
Built like the rest of Konstantly: secure, audited, license-aware
Plain-English access
Ask Claude, Cursor, or any MCP-aware agent about your workspace in normal language. The agent handles the rest under your account.
No SDK · No glue code · Works in any MCP client
Permission-aware by design
Every key inherits the issuing user's role. A learner key only reads learner-scoped data. An admin key can take admin actions. Revocation is instant.
Per-key scopes · Role hierarchy · License-tier aware
Strong keys, audited end-to-end
High-entropy keys, hashed at rest, shown only once. Every write action lands in your audit log with the same shape as the corresponding UI action, so compliance reporting just works.
Hashed at rest · Shown once · Audit log integrated
Rate-limited, brute-force resistant
Per-tier quotas keep usage predictable. Failed-auth attempts throttle per source IP so brute-force surfaces as alerts rather than access.
Per-key quotas · Per-IP throttling · Retry-After honored
Same engine as the UI
MCP isn't a parallel API. It runs the same business logic, the same permissions, and the same audit log as the browser experience, so behaviour stays consistent and supportable.
No re-implementation · No drift · Same SLA
Visibility you get for free
Every call records the tool name, duration, and outcome. Fleet-wide telemetry surfaces tool-selection patterns and failure spikes per tier, useful when rolling out to a team.
Per-tool metrics · Per-key history · Per-tier rollups
Generate courses from a prompt
Drop a PDF or describe the topic, and the agent builds a complete course with structure, lessons, and a final quiz. Grounded in your source materials, audited, ready to assign.
PDF / DOCX / images · Quiz included · Business and Enterprise
Start free. Ship your first course today.
free for 10 learners · no credit card