01
CORE AGENTIC LOOP
ProtoAgent streams model output, executes tools, appends results, retries transient failures, and keeps going until the model can answer directly.
the core of the coding agent
★BUILD-IT-YOURSELF AI CODING AGENTDIY AI CODING AGENT★MULTI-STEP TOOL LOOPSTOOL LOOPS★MCP + SKILLS SUPPORTMCP + SKILLS★SESSION PERSISTENCESESSIONS★SUB-AGENTS FOR CLEAN CONTEXTSUB-AGENTS★BUILD-IT-YOURSELF AI CODING AGENTDIY AI CODING AGENT★MULTI-STEP TOOL LOOPSTOOL LOOPS★MCP + SKILLS SUPPORTMCP + SKILLS★SESSION PERSISTENCESESSIONS★SUB-AGENTS FOR CLEAN CONTEXTSUB-AGENTS
PROTOAGENT
█▀█ █▀█ █▀█ ▀█▀ █▀█ ▄▀█ █▀▀ █▀▀ █▄ █ ▀█▀ █▀▀ █▀▄ █▄█ █ █▄█ █▀█ █▄█ ██▄ █ ▀█ █
A coding agent you can build yourself
Coding agents can feel like magic. ProtoAgent is a lean, readable implementation that pulls back the curtain, giving you the blueprint to understand and build your own.
Small enough to understand in a 20 minutes, simple enough to build yourself in an afternoon, usable enough to try out on an actual project.
// readable codebase
SMALL
// build it yourself
SIMPLE
// use it in real projects
USABLE
TRY IT OUT - LIVE SESSIONINITIALIZING...
Initializing session...
02
FILES, SHELL, AND WEB
It can read, write, and edit files, search a repo, run shell commands with approvals, fetch docs from the web, and keep a TODO list while it works.
taking action
03
APPROVE TOOL USES
Writes, edits, and non-safe shell commands go through inline approvals, while a small set of dangerous shell patterns stays blocked outright.
local safety rails
04
SESSIONS
Sessions are saved to disk with message history and TODOs, so you can quit, come back later, and keep working without starting from scratch.
persistent context
05
SKILLS AND MCP
Skills let you package project-specific instructions, and MCP lets you connect external tools without hardcoding them into the app.
easy extension points
06
SUB-AGENTS
Focused research can be pushed into child runs so the parent conversation stays lighter and easier to follow.
keep context clean