byte logo

byte

Free

Self-hosted AI agent — local LLM with tool calling, WASM sandbox, hybrid RAG on pgvector, MCP, and prompt-injection defenses

Visit Website
Related by shared tags and keywordsAmethystfixedyetotto_agent
ai-agentfastapilanggraphllmmcpaiai assistantautomation

Overview

Byte is a self-hosted local AI agent that runs an open-source model locally and can program, search the internet, and use tools via MCP without relying on paid APIs. The project states that phases 0 through 6 work end-to-end locally, including an agent with tools, a WASM sandbox, hybrid RAG, MCP, an OpenAI-compatible endpoint, multiuser support with JWT, and an interactive CLI, with 480 tests. Its stack includes Python, FastAPI, LangGraph, Ollama, PostgreSQL with pgvector, MCP, AG-UI, Pyodide/WASM, and Docker.

The architecture routes web or CLI requests through FastAPI to a LangGraph agent with a Postgres checkpointer, then to tools for web search via Tavily, a Pyodide WASM sandbox, hybrid RAG via pgvector, and MCP servers such as n8n. It can run without a database in memory for testing, but production mode requires Postgres; web search and code execution are optional depending on API keys and sandbox configuration. The project also provides an OpenAI-compatible endpoint, an interactive CLI, JWT-based multiuser support, repository navigation tools, and optional observability integrations.

Pricing

Free

Detailed plans have not been confirmed in our catalog. Check the official website for current limits and billing terms.

Visit Website

Prices and limits may change. Confirm the currency, billing period, seat minimum and usage caps on the official website.

Use Cases

  • Teams use Byte to run a local AI agent that can program, search the web, and use MCP tools without paid APIs.
  • Teams use Byte to execute code in a WASM sandbox via Pyodide.
  • Teams use Byte to search the web through Tavily when configured.
  • Teams use Byte to query documents with hybrid RAG using pgvector.
  • Teams use Byte to connect MCP servers such as n8n.
  • Teams use Byte to expose an OpenAI-compatible endpoint for OpenAI SDK clients.
  • Teams use Byte to navigate a repository with list_files, read_file, and grep.
  • Teams use Byte to manage multiuser access with JWT.

Features

Runs as a self-hosted agent

Uses a local LLM

Supports tool calling

Sandboxes execution with WASM

Combines hybrid RAG with pgvector

Connects through MCP

Includes prompt injection defenses

Reviews

No reviews yet. Be the first to share your experience!