Wire Maeve into the scripts and cron jobs you already have. Upload a folder of campaign media with one command, schedule a week of posts from a JSON file, or have a nightly job pull yesterday's numbers into your CI summary. Every command hands back clean JSON, and the few that do something irreversible refuse to run without --yes.
3 days free. Cancel anytime.
$ npm install -g maeve-cli $ maeve auth:login ✓ Approved in browser · token stored (7 days) $ maeve content:schedule --workspace ws_lilydia \ --content cnt_8f3k2 --at "2026-07-16T09:00:00+10:00" { "data": { "id": "cnt_8f3k2", "status": "scheduled" } } $ maeve content:publish --workspace ws_lilydia --content cnt_8f3k2 ✗ refused: this publishes immediately. Re-run with --yes
It installs from npm as maeve-cli, runs on Node 22 or newer, and the binary is maeve. Sign in through your browser for everyday work: auth:login opens an approval page, stores a revocable token for a week, and auth:logout kills it. For servers and CI, set MAEVE_API_KEY instead and skip the browser entirely.
Most commands name their workspace with --workspace, so a script is always explicit about which client, brand, or environment it is touching. Roles, plan gates, and provider limits are enforced on the server, so a script can never quietly do something a person could not.
Successful commands write JSON to stdout, the same envelope the API returns, so you can pipe results into jq, save them as a build artifact, or hand them to the next script. Failures write a structured JSON error to stderr with a code, a message, and sanitized details, then exit non-zero, so your scripts always know what happened.
The commands that go public or cannot be undone, publishing now, sending review links, public inbox replies, bulk media deletes, all refuse to run without --yes. A cron job cannot stumble into a destructive action by accident.
The practical, repeatable terminal work: list what is scheduled before standup, upload campaign media, create drafts from JSON, export a nightly analytics summary.
A command-line action is still Maeve work. Drafts, media, approval comments, inbox replies, reports, and grid items stay visible in the product surfaces your team already uses.
maeve-cli is the official command-line interface for Maeve. It lets teams inspect workspaces, integrations, drafts, schedules, media, approvals, inbox work, analytics, grid items, tags, and hashtags, then run supported write actions from a terminal, script, cron job, or CI job with the same API, role checks, and platform limits as the product.
Install it from npm with npm install -g maeve-cli. You can also use pnpm add -g maeve-cli, or run one-off commands with npx maeve-cli or pnpm dlx maeve-cli.
The binary is called maeve, so commands look like maeve auth:login, maeve workspaces:list, and maeve content:create --workspace <id> --json create-content.json.
maeve-cli requires Node.js 22 or newer. The package is distributed as an ESM Node CLI and is built with commander and zod.
The CLI supports browser login and API-key authentication. Browser login runs maeve auth:login, opens a Maeve approval page, and stores a revocable ezb_cli_ token locally. API-key auth uses MAEVE_API_KEY or --api-key, with the environment variable preferred for scripts and CI.
Use browser login for a person working from their own terminal. Use an API key for CI, cron, servers, or any headless environment. Avoid --api-key for routine use because command flags can show up in shell history or process listings.
Browser-login tokens are stored per API URL in %APPDATA%\Maeve\cli-auth.json on Windows, ~/Library/Application Support/Maeve/cli-auth.json on macOS, and ${XDG_CONFIG_HOME:-~/.config}/maeve/cli-auth.json on Linux. MAEVE_CLI_AUTH_FILE can override the location.
Browser-login tokens expire after 7 days. Device approval codes expire after 10 minutes if they are not approved in the browser.
Yes. Run maeve auth:logout to remove the stored token and revoke it server-side when possible. You can also revoke active CLI logins from the Developer settings page in the Maeve app.
Run maeve workspaces:list or maeve auth:whoami to see accessible workspaces. Most workspace commands require --workspace <id>. API keys are organization-scoped, so they can reach workspaces in that organization, but each request still goes through workspace access and role checks.
The CLI can read accessible workspaces, integrations, publishing capabilities and safe provider options, content, failed and approval counts, approval history, client review batches, media, media folders and tags, taxonomy, hashtag groups, analytics summaries, post analytics, demographics, inbox threads and messages, inbox stats, and grid planner items.
Yes. content:create can create drafts or content records from JSON, content:update can update supported fields, and content:schedule can schedule or reschedule an item. The same post appears in the Composer and Calendar.
Yes, content:publish can queue an immediate publish for a draft or scheduled content item, but it requires --yes. Scheduled posts created by the CLI still publish through the normal Maeve scheduling system and provider validation.
Yes, where the connected platform supports replies. inbox:reply sends a public reply and requires --yes. The current Social Inbox coverage is Instagram, Facebook, and Threads.
Confirmed --yes commands include content:publish, content:request-approval, content:resubmit, client-reviews:send, client-reviews:resend, inbox:reply, inbox:moderate, inbox:retry-message, inbox:delete-failed, inbox:read-all, media:bulk-delete, grid:delete, grid:promote, and analytics:report.
Yes. Successful commands write JSON to stdout, usually the same { data, meta } envelope returned by the API. Errors write structured JSON to stderr with error.code, error.message, optional status, and sanitized details, then exit non-zero.
Yes. Role checks are enforced server-side. Viewers can read supported workspace data but are blocked from mutations that require editor or admin access. Editors and admins can run write commands only where the product and plan allow them.
Yes, on workspaces whose plan includes those workflows. The CLI can read approval history, request or resubmit internal approval, submit decisions, add comments, and manage client review batches. Sending or resending review links requires --yes. See Approvals and agency workflows.
Yes. It can fetch analytics summaries, integration health, post analytics, aggregate post analytics, single-post analytics, demographics, and agency-plan PDF reports. analytics:report writes the PDF to --output and requires --yes.
No. The CLI is a terminal interface for humans, shell scripts, cron jobs, and CI. The REST API remains the lower-level surface for custom integrations and services.
No. The CLI is for terminals and scripts. MCP is for AI tools such as Claude, Codex, Cursor, and other MCP-compatible clients.
The CLI does not bypass Maeve roles, plan gates, media limits, scheduled-post limits, provider validation, inbox platform constraints, or API rate limits. It does not currently expose every product workflow, such as social-account connection, member or billing management, or Workbench planning-table commands.