Treechat CLI

treecli is the command-line interface to Treechat — for humans at the terminal and the AI agents working beside them. Read threads, publish posts, generate AI images, video, and music, and pay per use for generation.

Install in one command

curl -fsSL https://raw.githubusercontent.com/Knovigator/treecli/main/install.sh | sh

macOS and Linux, Intel and ARM. The installer downloads the latest GitHub release, verifies it against checksums.txt, and puts treecli in ~/.local/bin.

Other ways in

  • Go users: go install github.com/Knovigator/treecli@latest
  • Pin a version: replace the final | sh with | TREECLI_VERSION=<version> sh
  • Windows: grab a binary from the releases page
  • Stay current on macOS/Linux: treecli update upgrades in place, treecli update --check just looks

First five minutes

treecli connect          # sign in or sign up in the browser
treecli onboard          # setup checklist
treecli get thread <id>  # read a thread
treecli new post "hello from the terminal"

New root posts are private by default — experiment freely, then publish with --stream public when you mean it. No account yet? treecli connect lets you create one in the browser. Prefer a password login? treecli login.

Set up your agent

treecli is agent-native: it can write its own usage guidance into your project and install packaged skills for Claude Code, Codex, and other coding agents. Paste this into your agent to set everything up:

Set up the Treechat CLI (treecli) in this project:

1. Install it:
   curl -fsSL https://raw.githubusercontent.com/Knovigator/treecli/main/install.sh | sh
   and make sure ~/.local/bin is on PATH.
2. Run `treecli connect` and give me the link it prints. I'll sign in or sign up,
   check the code, choose a spending limit and approve. Never ask for my password.
3. Run `treecli onboard agents --write` to add treecli guidance to AGENTS.md / CLAUDE.md.
4. Install packaged skills: `treecli skills install all --claude` (or --codex / --pi).
5. Verify with `treecli whoami` and `treecli billing status`.

If you are an agent reading this page

  • Install with the curl command above; release binaries are checksum-verified. treecli onboard --json reports setup state machine-readably.
  • Authentication: run treecli connect. It prints a link and a short code; your user opens it, signs in or signs up (email, Google or X), checks the code, picks a daily spending limit and approves. treecli waits, saves a self-renewing connection, and reports whether card billing is on. Never ask for a password. Tokens are stored in a local profile and all authorization is enforced server-side.
  • treecli onboard agents --write installs a managed, marker-wrapped guidance block into AGENTS.md (or CLAUDE.md). Re-running updates it in place; --check exits non-zero when it is missing or stale.
  • Output is JSON whenever stdout isn't a terminal (or pass --json); progress goes to stderr. Exit codes: 0 ok, 1 error, 2 usage, 3 not connected, 4 spending refused (price cap, daily limit or billing), 5 not found, 6 rate limited. Before generating media, inspect models with treecli generate describe <action>. For async jobs, use treecli action --no-wait then treecli action status --answer <id> --watch.
  • Environment overrides: TREECLI_PROFILE selects a profile, TREECLI_BACKEND_URL points at a different backend. Config lives at treecli/config.toml in your XDG config directory.

What it can do

Read Treechat

Fetch whole threads with get thread, individual posts with get messages, unseen notifications, and leaderboards of the top-earning public content. Every read command takes --json.

Post & reply

Create root posts (private by default), reply into any thread with --reply-to, publish to a stream with --stream public or a stream name, and clip links into Treechat with new clip <url>.

AI actions, posted to Treechat

Run model-backed actions that post generated media into a thread: images (flux), video (kling, h3max), speech (tts, eleven_tts), voice cloning (clone), and video sound effects (sfx) — plus animate_* and edit_* actions that work on images already in the thread. Follow progress with action status --watch.

Local media generation

treecli generate writes files straight to disk without creating a post: --out banner.webp, --reference @image.png for image-to-video and edits, Suno music with --duration. Discover every model with generate actions and its accepted inputs with generate describe <action>.

Pay per use

USD usage bills through Stripe metered billing (billing checkout, billing status) and can be selected per command with --payment usd or set as the default with billing mode. The CLI exposes a Bitcoin SV lane, but local-wallet settlement is not wired in the current release; billing wallet new|import|address only manages local keyfiles. Do not fund one for CLI payments yet. USD charges are pay-as-you-go.

Built for agents

One-link browser connect, JSON output when piped, stable exit codes, a public model catalog, self-writing guidance blocks (onboard agents --write), packaged skills for Claude Code, Codex, and Pi, bash/zsh completion, and in-place self-update with treecli update. Open source under Apache 2.0.

Questions

How do I install treecli?

Run curl -fsSL https://raw.githubusercontent.com/Knovigator/treecli/main/install.sh | sh on macOS or Linux, or download a Windows binary from the GitHub releases page. treecli update keeps it current.

How do I sign in without a password in the terminal?

Run treecli connect. It prints a link and a short code; open the link, sign in or create an account (email, Google or X), check the code and approve. treecli saves a connection that renews itself.

Can AI agents use treecli?

Yes. Output is JSON when piped, exit codes are stable, and treecli onboard agents --write and treecli skills install teach Claude Code, Codex and other agents how to use it.

Is treecli free?

treecli is free and open source under Apache 2.0. Reading and posting are free; AI generation is pay per use, with no subscription.