Model Context Protocol (MCP)
Seamlessly extend agent capabilities by connecting standard MCP servers. Instantly query databases, local code indices, or external APIs.
Interactive Core Engine
Custom-PI is a self-evolving coding agent that never forgets, never stops, and operates entirely on local developer hardware with absolute privacy.
Ecosystem
Beyond coding, Custom-PI is packed with tools to automate your creative, social, and operational workflows securely on your local desktop.
Seamlessly extend agent capabilities by connecting standard MCP servers. Instantly query databases, local code indices, or external APIs.
Securely stores API keys, database credentials, and SSH keys locally in an AES-256 encrypted SQLite database with master key authentication.
Automates scheduling and publishing text, threads, and attachments directly to Twitter/X, Reddit, Bluesky, Discord, and Telegram using Playwright.
Features an interactive canvas editor for images (filters, crops, adjustments) and a side-by-side markdown document editor with AI co-writing.
Built-in Pollinations.ai and local SD webui pipeline. Generates batches of image assets and previews them before executing blog or social deployments.
Set hard spending limits for cloud LLM APIs. Track token consumption and real-time costs with automatic circuit-breaking caps.
Pipeline
An orchestration workflow that loops planning, reasoning, and tool execution to solve goals. Hover over each node to see its description.
Voice & Synthesis
Custom-PI integrates with open-source TTS synthesis engines like Kokoro and Piper via a Python subprocess. Wake words let you control code swarms hands-free, while real-time synthesis delivers auditory status reports.
Voice: Rachel • US Female
“Analyzing workspace. Ready to deploy the authentication changes.”
Setup Guide
A terminal-first installation process. Bring your own keys, connect your local repositories, and start building.
Add custom-pi to your workspace or run it globally. Installs fastify server and client UI bundle.
Plug in your Anthropic, OpenAI or Google API keys via vault, or launch completely offline using Ollama.
Access 30+ interactive panels. Plan task queues, review edits, stream TTS voice, and publish posts.
Also Available as a Desktop App
Standalone Electron app with system tray, auto-updates, and integrated web dashboard. No terminal required.
Engine Stack
A technology selection built to run with maximum efficiency on local developer environments, ensuring 0% latency and 100% code security.
Runs local server engines with filesystem I/O access.
Powers the responsive SPA dashboard and agent panels.
Ensures absolute code robustness across modules.
Handles high-speed local HTTP routes and system endpoints.
Saves vaults, settings, tasks, and memory graphs locally.
Streams tokens, log outputs, and TTS voice packets.