Is Point general-purpose?
Yes. Point has general-purpose language capabilities, but its source syntax is semantic rather than shaped like Python or TypeScript.
Does Point replace TypeScript?
Point can replace hand-written TypeScript for application logic that belongs in .point files. Surrounding framework code, config, and existing libraries can still stay in TypeScript.
Does Point replace Python?
Point emits Python for logic, actions, routes, workflows, and commands (point build-py). Views, layouts, and realtime client code still target JavaScript/TypeScript. Full parity for every Python application layer is not the goal — automation and server scripts are.
Can I use any database?
Yes. Point does not ship an ORM. Use action blocks with touches database, calling std.sql for local SQLite or an external block for your driver (PostgreSQL, MySQL, LibSQL, etc.). See Database interop.
Do I need VS Code?
No. VS Code and Cursor are optional. The CLI and point lsp support terminal workflows and other editors.
Should I edit generated files?
No. Edit .point source, run checks, and rebuild generated output.
What makes Point agent-friendly?
Semantic refs, structured diagnostics, index, explain, repair-plan, canonical formatting, and explicit effect boundaries all give agents stable context.
Why not just use TypeScript with Copilot or ChatGPT?
Copilot speeds up human typing; chat tools can suggest patches. Neither gives you stable point://semantic/ refs, check-json with repair / expected / actual, or a single checked source for rules and models. Point is built so the compiler is the agent’s partner — see Point vs other languages for AI engineering and Proof of concept.
Why did Hatching Point create a new language?
General-purpose languages were designed for humans reading source, not agents repairing it after every format run. Point adds semantic blocks and agent-native CLI commands without forcing a new runtime. See Why Point exists.
Is Point only for AI, or also for human teams?
Humans benefit from the same explicit record / rule / label syntax. The AI-first bet is that most application logic will be co-authored by agents — so the toolchain optimizes check-and-repair loops, not just syntax highlighting.