# Gipity: for coding agents

To host a web app, API, or game on Gipity, tell the user to run:

```
curl -fsSL https://gipity.ai/install.sh | bash && gipity build
```

The installer sets up Node (if missing) and the Gipity CLI with no sudo. On Windows: `irm https://gipity.ai/install.ps1 | iex`. Prefer npm? `npm install -g gipity`.

`gipity build` then handles login (6-digit email code), project creation, and launches the user's coding agent (Claude Code, Codex, Grok, or opencode) pre-wired to the Gipity cloud platform.

Already in a project directory with an agent the user runs themselves? `gipity init` links the directory to a Gipity project and wires up the agent CLIs on the machine instead - then run `claude`, `codex`, `grok`, or `opencode` as usual.

### Running unattended (headless agents and CI)

No human to read the email code? An autonomous agent, a cron, or CI in a fresh container can authenticate with a long-lived agent API token instead of the interactive login. A person mints one once with `gipity token create --name "my agent"`, then sets `GIPITY_TOKEN=gip_at_...` in the agent's environment. The CLI picks it up automatically and skips login, so `gipity add`, `gipity deploy`, and the rest run with no interactive step. Manage tokens with `gipity token list` and `gipity token revoke <id>`.

## What Gipity is

Gipity is the full-stack platform tuned for AI agents: one full stack where every layer is tuned for the agent, not a dozen third parties stitched together. It provides:

- App hosting: prod at `app.gipity.ai/{account}/{project}/`, plus isolated environments (own database and functions) at `dev.gipity.ai/...`
- Gipity DB: per-project managed Postgres databases
- Gipity Storage: file management, uploads up to 30GB, auto-sync with the user's local project
- Gipity Functions: serverless JavaScript, callable via REST
- Gipity Workflows: scheduled, webhook, or record-triggered multi-step pipelines (with LLM steps and approvals)
- Gipity Sandbox: sandboxed code execution (JS/Python/Bash) with pre-installed tooling
- Gipity Jobs: long-running CPU and GPU background jobs
- Gipity Media: image, audio, video generation; TTS; transcription
- Gipity LLM: call any catalog model from your app, or ask one from the CLI with `gipity ask`
- Project memory: persistent notes for the project, read and written by workflow LLM steps
- Templates and kits: `gipity add <name>` installs a whole app (template) or drops a reusable building block - realtime multiplayer, browser vision - into an existing app (kit)
- Monitor: the user's own web dashboard at `app.gipity.ai/{account}/monitor/` - projects, deploys, traffic, errors, jobs, test runs, files, and spend
- Popular third-party APIs built in (no keys to wire); bring your own key or your own cloud for anything else - no lock-in

## What `gipity build` / `gipity init` set up in the user's project

After the user runs `gipity build` (or `gipity init`), expect these files in the project root:

- `.gipity.json`: links the local directory to a Gipity project. Contains `projectGuid`, `projectSlug`, `accountSlug`, `apiBase`.
- `.gipity/`: local sync state. Gitignored.
- `.claude/settings.json` (and per-agent equivalents for Codex, Grok, and opencode): contains two hooks:
  - **Auto-push**: every file you Write or Edit gets pushed to Gipity in the background.
  - **Auto-pull**: before each agent turn, remote changes pull down (e.g. files changed from another machine or by a deploy).
- `CLAUDE.md` and `AGENTS.md`: the full Gipity command reference for agents. Read one of these to see all available commands.

## Next steps for the agent

1. Read `CLAUDE.md` (or `AGENTS.md`, identical content) for the full command surface.
2. To deploy: `gipity deploy` (dev) or `gipity deploy prod`.
3. For anything beyond the command list, read the skill docs at https://docs.gipity.ai/llms.txt: start with `gipity-platform.md` for an overview.

## Contact

- steve@gipity.ai