Skip to content

Pluma

CI Docs License: MIT Release

Local LLM chat in a single binary. Tavern card support, character editing, image generation, text-to-speech — no installer, no runtime to manage, sensible profiles so things work on the first try.

Install

curl -fsSL https://plumachat.app/install.sh | bash

Detects OS + arch, fetches the latest release tarball, verifies sha256, extracts to the current directory. Read the script before piping into bash — it's short and does exactly what it says.

Tarballs + signatures live on every tagged release. Grab the file for your platform:

Requires Go 1.22+ and Node 20+.

git clone https://github.com/guygrigsby/pluma
cd pluma
make build      # vite + go → ./pluma
./pluma

Either way: serves on http://localhost:8787, opens your browser, and the three-screen first-run wizard takes you the rest of the distance.

Get started → Releases → Repo on GitHub →

What's in the box

Area What ships
Chat OpenAI streaming + Anthropic native plugin · SillyTavern preset import · mobile-first UI with desktop sidebar · resizable rail · per-conversation persona
Characters Tavern Card v1/v2/v3 (PNG + JSON) · AI character creator · auto-portrait via SD · chub.ai browser · personas anchored in the system prompt
Models HuggingFace browser for GGUF + MLX · cancellable multi-file downloads · format auto-detect
Images A1111 + ComfyUI workflow JSONs · in-chat wand button · workflow library
Voice Any OpenAI-compatible /v1/audio/speech endpoint · voice library for codec engines (OmniVoice, XTTS) with reference-audio cloning
Security AES-256-GCM at rest · OS keyring for keys · WebAuthn passkeys · embedded Tailscale (tsnet) · host allowlist · trusted-proxy aware
Quality of life Hot-reload dev loop · auto-titled conversations · sampler family auto-pick · system-message hoisting · reduced-motion respected

Where to next