Developer prerequisites

A short list of everything you need installed and configured before you start Build your first enterprise AI app or any other developer-track tutorial.

Local environment

Optional but useful: Visual Studio / Rider / VS Code with C# extension, Postman or curl, jq for inspecting JSON responses.

Curiosity packages

• Curiosity.Library (NuGet) — the SDK used to build data connectors and to talk to a Workspace from external code. Available on nuget.org.
• Curiosity.Library.Python — Python helper library if you prefer to write connectors in Python. Mirrors a subset of the C# SDK.

Custom endpoints and AI tools are written inside the Workspace (the C# is compiled by the workspace itself), so you don't need to add packages for those.

Credentials and tokens

• An admin password for your local workspace. Generate one and pass it via MSK_ADMIN_PASSWORD on docker run so the default admin/admin is never used. See the Configuration reference.
• An API token for ingestion (created from the UI: Settings → API Tokens). Scope it to ingestion only — never use an admin token from a connector. See Token scopes.
• A model-provider API key if you'll enable embeddings or chat:
  • OpenAI (sk-...) or Azure OpenAI deployment.
  • Anthropic (sk-ant-...).
  • Or a local server speaking the OpenAI-compatible API. See LLM Configuration for the per-provider setup.

Sample data

Pick the dataset that matches the tutorial you're following:

• Default tutorial — a small CSV of support tickets. The Build your first enterprise AI app page links to a starter set; any 50–500 rows with id, customer, product, status, subject, body, created_at is enough.
• Technical Support tutorial — clone curiosity-ai/technical-support.
• HackerNews tutorial — the connector fetches data live; no setup needed beyond outbound internet.

Network and ports

• TCP 8080 free on 127.0.0.1 (or pick a different host port with -p 127.0.0.1:9000:8080).
• Outbound HTTPS to your model provider (OpenAI, Anthropic, Azure) if you're using one.
• Outbound HTTPS to hub.docker.com to pull the image, and nuget.org to restore packages.

Access to the documentation

Keep these pages open while you work:

• API Overview — what's reachable from a connector or endpoint.
• Graph Query Language — the Q() chain.
• Search DSL — SearchRequest shape.
• Configuration reference — every MSK_* env var.
• Troubleshooting — common failures and fixes.

Ready check

Before you start the tutorial, run:

docker --version
dotnet --version
docker pull curiosityai/curiosity:latest

All three should succeed. Then continue with Build your first enterprise AI app.