Download

A private, offline AI system for schools and teams

Your class, your model.

An in-house AI system for schools and teams. It runs on your own hardware, answers from your own documents and lists its sources. Pair it with a local model and your questions and documents never leave the building.

See what it can doSecurity overviewGitHub

Runs with

  • Ollama
  • LM Studio
  • LocalAI
  • KoboldCPP
  • Docker
  • Kubernetes
  • Helm
  • OpenShift

What it's for

One system for classrooms, students, courses, instructors and teams.

A student's question about carrying capacity in a course workspace, answered from the readings, with each reading listed under the answer.

A course assistant for the whole class. It answers from the course's own readings and lists the sources under each answer.

More about classrooms

Study help drawn from the assigned material. Ask for an explanation, a summary or practice questions, and keep each topic in its own thread.

More about study help

Build a workspace for each course from the syllabus, readings and handouts, add your students, and see how they use it.

More about teaching a course

Teach faculty modern AI workflows, agentic coding included, on a local model: agents, the skills you turn on and MCP servers, all running on your own hardware.

More about instructor training

Put policies, manuals and research in workspaces, decide who sees which, and get answers with the sources listed. Each workspace picks its own model.

More about teams and labs

Grounded answers

Answers from your own documents, sources listed.

When a workspace has documents, Mission LLM searches them on your server before the model answers. Open Sources to see which documents an answer used, the passages it pulled and how closely each one matched.

  • Sources listed under grounded answers
  • Retrieved passages with match scores
  • Query mode, with a refusal message you write

How your data flows

A student's answer about energy flow and the nitrogen cycle, with the Sources panel open: each course reading it drew on, a passage from it and its match score.

Security and control

You set the boundary.

The Privacy & Data-Handling page with the anonymous telemetry switch off, and the vectors kept in the built-in LanceDB store on the instance.

Set DISABLE_TELEMETRY="true" before the first start and no usage events are sent. Or turn it off in Settings.

Learn more about telemetry

Pair it with a local model runtime and stage the built-in models. Chat, embeddings and search then run with no connection at all.

Learn more about offline use

Give instructors or team leads the manager role, which opens every workspace. Everyone else sees only the workspaces they are added to.

Learn more about access

Admins and managers can read every workspace chat, see who sent it, and export it as CSV, JSON or JSONL.

Learn more about oversight

Sign-ins, failed sign-ins, and changes to users, API keys, invites, workspaces and documents, in one log for admins.

Learn more about oversight

Up and running in four steps.

  1. Install on your server

    One container, with every document and chat in one storage folder on your hardware.

  2. Connect a local model

    Point it at Ollama, LM Studio or another runtime on your own network, with no cloud account. Approved cloud providers work too.

  3. Add your documents

    Create a workspace for each course, cohort or team. Add its readings, handouts or manuals, or connect a code repository. Mission LLM indexes them on your server.

  4. Invite your people

    Turn on multi-user mode, invite students, instructors or team members, and add each one to the workspaces they need.

In-house

Run it on a lab workstation, a school server or a rack of your own.

Mission LLM ships as one container that serves the app and API on port 3001. Start it with Docker, Compose, Kubernetes, Helm or OpenShift, on your own hardware or, if you prefer, in a cloud account you control.

See every install option

Docker on Linux or macOS
export STORAGE_LOCATION=$HOME/missionllm && \
mkdir -p $STORAGE_LOCATION && \
touch "$STORAGE_LOCATION/.env" && \
docker run -d --rm -p 3001:3001 \
--cap-add SYS_ADMIN \
-v ${STORAGE_LOCATION}:/app/server/storage \
-v ${STORAGE_LOCATION}/.env:/app/server/.env \
-e STORAGE_DIR="/app/server/storage" \
REGISTRY/mission-llm:latest
REGISTRY is a placeholder. Public images are not published yet. Today you build the image from source (see Download).

Resources

MIT licensed foundation.

Open at the core. Yours to run.

Mission LLM is built on the open-source AnythingLLM project (MIT License) and adds an in-place upgrade, a documented list of outbound connections and deployment templates that pull only your image.

License
MIT
User roles
3
Cloud templates
3

See the AnythingLLM project

What Mission LLM adds, on top of the AnythingLLM core

  1. In-place upgrade from AnythingLLM

    The database, settings and vector collections carry over.

  2. Documented outbound connections

    Every connection the server can make, and how to avoid it.

  3. Templates that pull only your image

    Deployment templates run the image you build and host.

AnythingLLM core (MIT License)

  • Ingestion
  • Retrieval
  • Agents
  • Providers

Bring it in-house.

Start with one container, a local model and one workspace: a course, a training cohort or a team. Questions first? Tell us what you teach or run and we will reply by email.

Talk to us