Local AI Assistant models (LLMs)
Works with Ollama-compatible models you install: general, coder, and specialist tags. Setup lists Pea-friendly options; your ollama list is what Pea can use. Not locked to one vendor.
Beta
PeaAI.org · Local LLMs · Desktop agent · 2026
Your PC.
Pea AI Coder
A real coding agent on your machine: natural chat, pea-tools file edits, websites, apps, games, and Linked folders. Models run locally through Ollama (any Local AI Assistant / LLM you install). Writes hit disk only after Keep.
Pea is in active development. Expect changes — and tell us what you want next on Discord.
0%
Runs on your PC
Keep
Keep before write
Soon
Public download soon
Pea AI Coder · workspace
// Local AI Assistant · min. 6 GB VRAM · Keep gated
import { refactor } from './api';
const ok = await refactor('src/App.jsx');
// …
if (ok) pea.showDiff();
"Review → Keep"
Product
What Pea AI Coder actually does
Same desktop app you download: explorer, chat, editor, preview, review bar, and a Python backend talking to your Local AI Assistant runtime.
Keep before write
propose_edit and patches land on disk only after you Keep. Diff, review, undo. Shell and uploads need confirm.
Live HTML preview
Open HTML beside the editor and watch it update as Pea edits. Open in browser when you need linked CSS and JS.
Agent tools + PEA.md
todo_write, read/list/grep, str_replace, web_search/fetch_url, check_env, run_shell. Project memory stays in PEA.md across turns.
Hardware-aware context
CPU, RAM, and NVIDIA VRAM awareness. Context profiles tuned so Pea stays usable on mid GPUs (min. 6 GB VRAM for comfortable coding).
Websites, apps, games, chat
Premium web UI with named design styles, app scaffolds, Unity/Godot/Unreal/web game layouts, and natural chat when you are not coding. Full local development freedom across the stack.
Build
What people use Pea for
Real project work on a private machine, not a generic chat demo.
Websites and UI
Landing pages and tools with atmosphere, motion, and working controls. Style vocabulary when you name it (glassmorphism, brutalism, liquid glass, and more).
Apps and linked folders
Desktop and web apps with real entry points. Linked workspace roots, deep links, and shared contracts when pieces connect.
Games
Engine-correct layouts for Unity, Godot 4, Unreal 5, or web loops. Playable path before cosmetics; docs lookup when APIs change.
Full local freedom
Private by default. Full creative and technical freedom on your machine: you decide the stack and the brief. What you build stays between you and your PC.
Why Pea
Why not just a browser chatbot?
Your machine
Default is local Ollama on 127.0.0.1. No required Pea cloud account for coding.
Transparent writes
Every file write goes through Keep. You see the change before it hits disk.
Model freedom
Install the Local AI Assistant tags you want. Pea routes and profiles around what you have. Your machine, your rules.
Desktop workspace
Electron app with explorer, chat, editor, preview, and review, not another browser tab.
Community
Built in the open — with you
Pea is still Beta and under continuous development. Ideas, bug reports, and feature wishes are welcome on Discord.
Beta
Discord is the home for feedback
Share what you build, request tools, vote on priorities. No corporate support queue — a small indie community shaping Pea.
Support
Keep Pea independent
Pea is built independently. Paid tips will return when checkout is ready. No ads, no cloud lock-in.
Coming soon
Tips are not open yet
Card checkout is not live. Join Discord for updates. When support opens, this section will use a proper payment provider and updated legal terms.
FAQ
Common questions
Straight answers for searchers and people deciding if local coding with Pea is for them.
What is Pea AI Coder?
A local Windows desktop coding agent. You chat, it proposes file edits, and writes reach disk only after Keep. Models run on your PC through Ollama.
Does Pea send my project to the cloud?
Pea is built to run locally with Ollama on your machine. This website is static and does not host your projects. Public download is not open yet.
When can I download Pea AI Coder?
The public Windows package is coming soon. Requirements and install steps are on the Download page so you can prepare hardware.
What hardware do I need?
Windows 10 or 11 64-bit and at least 6 GB VRAM for comfortable local coding. More VRAM allows larger models. Disk space depends on the models you install.
Where do I get updates and give feedback?
Join the Discord community linked on PeaAI.org. Paid tips and installer files are not published yet.
Ready to run Pea on your PC?
Download the Windows package, run Setup, pick Local AI Assistant models, launch Pea AI Coder.
Download for Windows