How it works
Connect the pieces, strengthen the whole.
Most AI tools give you an answer and stop there. Quilt does the work, and your district decides, in advance, how far it can go.
The foundation
One foundation for your K–12 agents.
Connect your systems once. Define your district’s rules. Use that shared foundation to run Quilt’s prebuilt agents or build your own, without rebuilding access controls for each workflow.
Retrieve the information you need
Connect district systems, documents and public sources through supported APIs, standards and approved files.
Built with K–12 experience
Agents shaped by decades of experience from district leaders and operators, with workflows designed for school systems.
Start with practical district work
Analyze data, flag changes, draft documents, find policy answers and check sources with specialized agents.
Keep your district in control
Set what people and agents can access, what they may do and which steps require staff approval.
How the pieces fit
Your systems remain the source.
Quilt connects the systems you already use with the context that makes your district unique. Your documents, strategy, policies and goals. Together, they give K–12 agents the information and direction they need.
- K–12 agentsPrebuilt and custom agents analyze information, monitor changes and prepare work for review.
- Quilt infrastructureConnects to approved sources, retrieves information, applies district context and coordinates agents.
- Foundation · Existing systemsSIS, HRIS, data and assessment platforms, data warehouses, finance and learning systems.
- Foundation · District contextInternal documents, policies, district strategy, LCAP, improvement plans and local definitions.
- District governance · Every layerAccess permissions, allowed actions, required approvals and activity records apply across the architecture.
Available through chat, APIs and MCP.
Dashboards show numbers. Agents do the work.
Most education AI is still a dashboard or a copilot. A dashboard shows you numbers. A copilot helps you write. Neither one does the work.
The shift worth watching is toward agents that work at the level of the institution. They retrieve information, reconcile systems, calculate, draft, monitor and route work, and eventually run the workflows your district has approved.
- Summarize a district policy
- Draft a lesson plan
- Answer a question about student data
- Help write an email
- Review student performance across multiple data sources
- Identify students who may need support
- Compare results against district goals
- Create an intervention plan or report
- Prepare individualized communications
- Take approved actions in connected systems
A chatbot is something you ask.
An agent is something you can delegate to.
And because this is K–12, districts stay in control. Administrators determine what each agent can access, what it can do, and which actions require human approval.
Step 1 · What an agent draws on
Three inputs behind every answer.
A chatbot has one source, the internet. A Quilt agent has three, and shows you all three every time it answers.
What’s happening
Enrollment, attendance, grades, assessments, schedules, staffing. An agent only reaches the records the person asking is already allowed to see.
What your district decided
Board policy, administrative regulations, improvement plans, MTSS guidance, handbooks and site plans. Every page carries who may read it.
What everyone else knows
Education Code, CDE guidance, Dashboard files, peer practice and research, searched in an isolated sandbox that never receives a student’s name.
You can see these three lanes on every answer in the demo on the homepage.
Step 2 · What happens in between
Most of this isn’t the language model.
The model occupies two steps of ten. Everything that makes this safe for student data is ordinary, auditable software, which is why it holds when the models change.
Who you are, your roles, your org scope and your relationships. All resolved from your identity provider, not from what you typed.
The authorization engine builds a restricted tool set for this person and this request.
Sensitivity, complexity, latency and modality select an approved provider.
The agent requests data through the tools it was given. It can’t reach anything else.
Rates, growth and thresholds are calculated in the metric layer using your definitions.
The model explains what the numbers say and labels fact, calculation, reading and recommendation.
Privacy and citation validation. A claim without a supporting source doesn’t ship.
Step 3 · Who decides what an agent can do
Your administrators set the boundary, not us.
Every agent has three separate settings, and all three are yours. Nothing expands because we shipped a release.
What it can access
Which systems, which fields, which students, inherited from the roster and relationships already in your SIS, not re-created in a settings page.
What it can do
Read, analyze, draft, prepare. Writing to a system of record is off by default and stays off in year one.
What needs a human
Which actions stop for approval, and who approves them. Some (discipline, eligibility, placement, retention) can never be turned on at all.
Step 4 · Who does the work
You never pick an agent. You ask a question.
Quilt routes the question to whichever agents hold that corner of the work, then tells you which ones answered so you can check them.
Computes every rate, average and trend straight from the records, so the math is checkable rather than generated.
Watches for students slipping before a report card says so, and tells you which week it started.
Writes the letters, memos and board decks, each one specific, none of them sent without you.
Checks every claim against the record it came from and labels what’s a fact versus a reading.
Knows your board policy, curriculum guides and state rules, and cites the page.
Point Quilt at a workflow your team repeats and it becomes another agent, bound by the same limits as the rest.
Meet the team
