> For clean Markdown of any page, append .md to the page URL. > For a complete documentation index, see https://docs.aventure.vc/llms.txt. > For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.aventure.vc/_mcp/server. # CLI quickstart The CLI wraps the same [API](/api-reference) and the same [credentials](/authentication). Public reads work with no credential at all, so you can run a real query before you sign in. The CLI requires Node.js 24.18 or later in the 24.x series. `npm install` fails on any other major version. #### Install ```bash npm install --global @aventurevc/aventure-cli --registry=https://registry.npmjs.org/ ``` The binary is `aventure`. #### Run a public read No credential required: ```bash aventure entities list --size 3 ``` Companies, products, and services come back as a table. Add `--json` for the full envelope or `--data` for the response body alone. #### Sign in ```bash aventure auth login ``` This uses the OAuth device authorization grant: it prints a one-time code and a sign-in URL, opens a browser when one is available, and polls until you approve. The same command works on a laptop, over SSH, and inside a container — `--no-browser` only skips the automatic open. The credential is stored in your operating-system keyring, or in `~/.config/aventure/credentials.json` with mode `0600` when no GUI session makes a keyring available. Access tokens renew from the refresh token automatically. Prefer a personal API key? `aventure auth login --key` creates and stores one through the same browser approval. An existing key can always be supplied through the `AUTH_TOKEN` environment variable instead — that is the path for CI and other non-interactive environments. #### Verify ```bash aventure auth doctor ``` `doctor` runs every environment check in one call and reports `{ ok, check: [...] }` covering the CLI version, the installed schema contracts, the API host, which credential scope resolved locally, and whether the command catalog matches the operation set the API serves. It exits `0` only when no check failed; a skipped check does not fail the run, and each failing message names the command that fixes it. `aventure auth status` is the narrower version — it reports credential provenance and sign-in providers, never the secret itself. ## The command shape `entities`, `people`, and `news` are the three record types, and each takes the same verbs, so learning one teaches the rest: ```bash aventure entities list --size 3 aventure entities get --entity-slug calendly-atlanta-ga-us aventure entities search --query "fintech companies in Austin" --size 3 aventure lookup --token e9TwndP056X3 ``` | Verb | What it does | | -------- | --------------------------------------------------------- | | `list` | Read many records, with `--page` and `--size` | | `search` | Find many from a plain-English `--query` or a filter set | | `get` | Read one record | | `lookup` | Resolve a name, URL, or domain to one record of that type | Each record type names its own slug flag — `--entity-slug`, `--person-slug`, `--news-slug` — so the flag tells you which record you are asking for: ```bash aventure entities get --entity-slug calendly-atlanta-ga-us aventure people get --person-slug yin-wu aventure news get --news-slug 2026-09-16-pulley-a-carta-rival-is-shutting-down ``` Bare `--slug` on these commands is the request **body** field used by `--batch`, not the single-record selector. Passing it alone returns [`400` or `422`](/errors). Use the owner-qualified flag above for one record, and `--batch --slug ` for many. Add `--batch` to `get` or `lookup` to resolve many records in one request. `aventure search` and `aventure lookup` are the cross-record versions: `search` runs companies, people, and news together, and `lookup --token` resolves any identifier — a public handle such as `e9TwndP056X3`, a UUID, a slug, or an external registry id like a ticker or EIN — to whichever record owns it. The first two commands read public data with no credential. `aventure lookup` needs a credential, and `aventure search` and `aventure entities search` additionally need a Pro subscription or trial — without one they answer [`402`](/errors). Run `aventure auth login` first. ## Output and exit codes Each command accepts an output mode: * `--text` — the default in a terminal: a table for a list, labelled fields for one record * `--data` — the response data as JSON, the default when piped or non-TTY * `--json` — the full envelope, capped at 50KB * `--data-full` — uncapped JSON, for bounded programs Exit code `0` means `envelope.ok === true`; `1` means the call failed. Scripts can branch on that without parsing output. ## Find the command for an operation `aventure docs api` prints what an operation needs without calling the API, so you can go from an [API reference](/api-reference) `operationId` to the command that runs it: ```bash aventure docs api getEntityDetail ``` ``` cli: aventure entities get GET /v1/entities/detail operationId: getEntityDetail resource: entities intent: read scope: read mutates: false destructive: false mcpTool: aventure_read body: none required path flags: none required query flags: none ``` `mcpTool` names the [MCP](/mcp) tool that reaches the same operation, and `scope` tells you whether a read or write credential is needed before you run it. `--help` on any command lists that command's own flags, grouped by whether they land in the path, the query string, or the request body. ## Next * [Answer a real question](/first-answer) — these commands chained into one result * [Authentication](/authentication) — which credential each operation accepts * [MCP quickstart](/mcp) — the same data inside an AI client * [Pagination and filtering](/pagination-and-filtering) — the filters behind `--size`, `--page`, and the rest * [`@aventurevc/aventure-cli`](https://www.npmjs.com/package/@aventurevc/aventure-cli) on npm, and its [source](https://github.com/aventurevc/aventure-cli)