Awesome AI AgentsAI Agents

bravenewxyz/agent-c

⭐ 442 C repository created 2025-08-25

Agent-C is a minimal terminal agent implemented as one C source file. It connects to the OpenRouter API with an API key taken from an environment variable, sends the conversation together with the definition of a single shell tool, and streams the model reply token by token using server-sent events so output appears as it is generated. When the model answers with a tool call, the agent runs the requested shell command locally, sends the captured output back as a tool result and continues the turn, allowing up to five tool calls before the turn ends. Conversation state is a sliding window of the last twenty messages including tool call history, which bounds memory use and prompt size without any external storage. The distinguishing property is the absence of dependencies: there is no libcurl and no OpenSSL, and transport security is handled by the platform, using SecureTransport on macOS, so the program builds with a plain make invocation and produces a binary of roughly four kilobytes after LZMA self-extraction on macOS or UPX compression on Linux. Both macOS and Linux targets are supported and auto-detected by the makefile. A separate specification document records the architecture and a set of test cases, which makes the repository readable as a worked example of how a tool calling agent loop works at the protocol level, without a framework in between. The code is released under CC0, so it can be copied without attribution requirements. It is aimed at developers who want an agent they can run on constrained systems, who want to inspect the raw request and response flow of an OpenAI-compatible tool calling API, or who want a starting point for embedding an agent loop in a C program.

https://github.com/bravenewxyz/agent-c

agentai-agentopenroutercclishelltool-callingstreamingminimalzero-dependency

Also in AI Agents

OpenHands/OpenHands

OpenHands is an AI-powered platform that automates software development tasks, enabling developers to code less and produce more by modifying code, running commands, browsing the web, and integrating APIs efficiently.

usestrix/strix

Strix is an open-source AI-driven security testing platform that autonomously finds, validates, and helps fix application vulnerabilities through dynamic and collaborative AI agents.

aaif-goose/goose

Goose is an open-source, native AI agent available as a desktop app, CLI, and API, designed for diverse tasks like code, research, writing, and automation, supporting over 15 LLM providers.

KeygraphHQ/shannon

Shannon is an autonomous AI pentester for web applications and APIs, analyzing source code, identifying attack vectors, and executing real exploits to prove vulnerabilities.

wshobson/agents

A comprehensive collection of 44 specialized AI subagents for Claude Code that enhance development workflows with domain-specific expertise across software development, infrastructure, security, data, and business domains.

agentscope-ai/agentscope

AgentScope is a transparent, modular, and customizable framework for building LLM-empowered multi-agent applications with real-time control and asynchronous execution.

iOfficeAI/OfficeCLI

OfficeCLI is an open-source command-line interface and API enabling AI agents to read, edit, and automate Word, Excel, and PowerPoint files without needing Office suite installations.

langchain-ai/deepagents

Deep Agents is an open-source agent harness built on LangChain and LangGraph that enables AI agents to plan, execute, and delegate complex long-horizon tasks using customizable tools, subagents, and human-in-the-loop workflows.