GitDiagram
Browse

AI agent architecture examples

Interactive architecture diagrams of 244 open-source repositories with a GitDiagram architecture diagram where the diagram has an LLM or AI-model component and a component or group named as an agent, most stars first. Each diagram shows the project's main components, how they connect, and links every part to its file on GitHub.

Across the 200 most-starred AI agents here, the median diagram has 22 components. 77% draw a data store (most often a vector store, files on disk and SQLite), 48% have a command-line interface and 26% a web interface.

100% have an LLM or AI-model component, 39% speak MCP, 27% have a plugin or extension system and 30% are split into workspace packages (packages/, apps/, crates/ and similar).

The files their components link to most are index.ts (34), main.py (27), app.py (26), server.ts (22) and App.tsx (20).

Group names that recur: “agent runtime” (32), “user interfaces” (20), “agent execution” (12), “agent capabilities” (10) and “platform services” (8).

By language: Python 42%, TypeScript 33%, Rust 8%, Go 5% and Jupyter Notebook 5%.

In the 200 most-starred diagrams
Has an LLM or AI-model component100%
Draws a data store77%
Has a command-line interface48%
Speaks MCP39%
Split into workspace packages30%
Has plugins or extensions27%

Repositories 1–50 of 244

  1. openclaw/openclaw★ 394k · TypeScript

    OpenClaw is a self-hosted assistant whose Gateway coordinates sessions, agent execution, tools, and messaging channels; users reach it through the Control UI, CLI/TUI, or…

  2. nousresearch/hermes-agent★ 261k · Python

    Hermes Agent is a multi-surface AI agent: users interact through the CLI/TUI or messaging gateway, the agent builds context and runs model/tool turns, and results return through…

  3. anomalyco/opencode★ 225k · TypeScript

    OpenCode is an open-source AI coding agent with terminal and desktop interfaces.

  4. ultraworkers/claw-code★ 195k · Rust

    Claw Code is the canonical Rust implementation of the claw CLI agent harness; the Python tree is described as a companion/reference workspace, not the primary runtime.

  5. instructkr/claude-code★ 195k · Rust

    Claw Code is a Rust CLI agent harness; the README identifies rust/ as its canonical runtime and src/ as a Python reference workspace.

  6. firecrawl/firecrawl★ 192k · TypeScript

    Firecrawl is a web-data API for AI agents: clients request search, scraping, discovery, and asynchronous jobs, then receive web content in LLM-ready formats.

  7. ollama/ollama★ 185k · Go

    Ollama is a local-and-cloud open-model runtime: users can run or chat with models through the CLI, call a REST API, or use the desktop app; integrations can also connect through…

  8. langgenius/dify★ 160k · Python

    Dify is an LLM application development platform.

  9. langchain-ai/langchain★ 148k · Python

    LangChain is a Python framework for composing interoperable LLM applications and agents.

  10. shubhamsaboo/awesome-llm-apps★ 140k · Python

    Awesome LLM Apps is a curated collection of independent AI-agent, agent-skill, and RAG applications—not one integrated runtime.

  11. openai/codex★ 133k · Rust

    Codex CLI is a locally running coding agent: a user starts a session, the agent works with a model and can use local tools, then presents progress and results in the terminal.

  12. browser-use/browser-use★ 120k · Python

    Browser Use is a Python AI browser agent: a user supplies a task and model, the agent plans and invokes registered browser actions, and returns a result with history.

  13. earendil-works/pi★ 119k · TypeScript

    Pi is an extensible coding-agent harness with interactive, print/JSON, RPC, and TypeScript-SDK entry points.

  14. juliusbrussee/caveman★ 108k · Go

    Caveman reduces tokens in AI coding workflows: an agent’s requests pass through a local proxy, which compresses eligible content before forwarding it to a provider, while…

  15. nexu-io/open-design★ 97.7k · TypeScript

    OpenDesign is a local-first, agent-native design studio: a user starts with a brief, an agent runtime works on project files, and the daemon exposes generated artifacts for…

  16. openhands/openhands★ 91k · TypeScript

    Agent Canvas is a self-hosted developer control center for coding-agent conversations and automations across local, remote, and cloud backends.

  17. 666ghj/mirofish★ 84.2k · Python

    MiroFish is a web application for running LLM-powered, multi-agent social simulations from uploaded source material.

  18. lobehub/lobehub★ 82.8k · TypeScript

    LobeHub is a workspace for building and operating AI agents, with chat and team collaboration as central user workflows.

  19. datawhalechina/hello-agents★ 81.9k · Python

    Hello-Agents is a Chinese open-source tutorial for learning to design and build AI-native agents, not a single runtime application.

  20. ruvnet/ruflo★ 80.8k · TypeScript

    Ruflo is an agent meta-harness for Claude Code, Codex, and MCP clients.

  21. abi/screenshot-to-code★ 79.1k · Python

    Screenshot-to-Code is a local/self-hostable AI application that converts screenshots, mockups, Figma-like inputs, URLs, imported code, and recordings into editable web…

  22. headroomlabs-ai/headroom★ 74.5k · Python

    Headroom is a local context-optimization product for AI applications and agents.

  23. usestrix/strix★ 74.3k · Python

    Strix is an AI-driven penetration-testing product: users launch local scans or use its cloud interface, agents explore targets and validate vulnerabilities, and the system…

  24. cline/cline★ 70.2k · TypeScript

    Cline is a coding-agent product spanning a terminal CLI, shared SDK/core, VS Code integration, and Cline Hub.

  25. code-yeongyu/oh-my-opencode★ 70k · TypeScript

    OmO is a multi-model agent harness for turning a user’s task into completed work.

  26. openinterpreter/openinterpreter★ 68.4k · Rust

    Open Interpreter is a Rust coding-agent CLI that emulates multiple agent harnesses, with a terminal interface and compatibility protocols.

  27. mem0ai/mem0★ 67.7k · Python

    Mem0 is a memory layer for AI applications, delivered as Python/TypeScript SDKs, a self-hosted API and dashboard, command-line tools, and agent-framework integrations.

  28. mintplex-labs/anything-llm★ 67.5k · JavaScript

    AnythingLLM is a multi-user AI application for chatting with documents, configuring model providers, and using agents.

  29. asgeirtj/system_prompts_leaks★ 67.4k · Python

    This is a static, community-maintained archive of captured AI system prompts and related agent/skill artifacts, not a running application.

  30. zhulinsen/daily_stock_analysis★ 65.7k · Python

    Daily Stock Analysis is a multi-market AI stock-research and decision-support product, delivered through a web/desktop workspace, API, and chat bots.

  31. jingyaogong/minimind★ 65k · Python

    MiniMind is a from-scratch, compact language-model project: its main user journey is preparing training data, training or adapting a model, then evaluating or serving generated…

  32. virattt/ai-hedge-fund★ 63.6k · Python

    AI Hedge Fund is an educational research and trading-simulation application.

  33. alishahryar1/free-claude-code★ 62.5k · Python

    Free Claude Code is a local model gateway and coding-agent environment: users configure providers and a model, launch a client, and send requests through a Claude-compatible API…

  34. crewaiinc/crewai★ 59.5k · Python

    CrewAI is a Python multi-agent automation framework: developers define role-based Crews or event-driven Flows, which coordinate model calls and can combine precise workflow…

  35. lordog/dive-into-llms★ 55.4k · Jupyter Notebook

    This repository is a free, hands-on educational tutorial series, not an application with a shared runtime.

  36. gpt-engineer-org/gpt-engineer★ 55.1k · Python

    gpt-engineer is a Python coding-agent experimentation platform: it turns a natural-language project prompt into generated code, can improve an existing codebase, execute…

  37. run-llama/llama_index★ 52.3k · Python

    LlamaIndex OSS is a Python framework for grounding LLM applications in user data, with retrieval-augmented querying and agent orchestration as central capabilities.

  38. hkuds/nanobot★ 49.5k · Python

    nanobot is a self-hosted personal AI agent framework.

  39. keygraphhq/shannon★ 48.1k · TypeScript

    Shannon is a local, autonomous AI pentester for web applications and APIs.

  40. alibaba/open-code-review★ 47.4k · Go

    OpenCodeReview is an AI code-review product exposed through a CLI and editor integrations.

  41. tashfeenahmed/freellmapi★ 47.2k · TypeScript

    FreeLLMAPI aggregates free and custom model providers behind an OpenAI-compatible API, with a router selecting available models and tracking per-key limits.

  42. librechat-ai/librechat★ 46.6k · TypeScript

    LibreChat is a self-hostable AI chat application supporting multiple model providers, Agents and tools, file workflows, and chat features such as search, artifacts, and audio.

  43. zhayujie/cowagent★ 46.1k · Python

    CowAgent is a Python-based, extensible AI-agent runtime: it receives multimodal messages, plans and executes tool-using tasks, maintains memory/knowledge, and replies through the…

  44. astrbotdevs/astrbot★ 45.7k · Python

    AstrBot is an extensible conversational-agent platform that connects messaging channels and Web Chat to AI providers, agent tools, plugins, and knowledge features.

  45. danny-avila/librechat★ 44.5k · TypeScript

    LibreChat is a self-hosted multi-model conversational platform with chat, Agents, Assistants, tools, files, MCP, artifacts, and automation.

  46. badlogic/pi-mono★ 44.3k · TypeScript

    This is a TypeScript monorepo organized around one core product, the Pi coding agent, plus shared libraries underneath it.

  47. anthropics/claude-plugins-official★ 43.9k · Python

    This repository is a curated Claude Code plugin directory, not one unified application.

  48. vectorize-io/hindsight★ 41.7k · Python

    Hindsight is a long-term memory system for AI agents.

  49. vectifyai/pageindex★ 38.9k · Python

    PageIndex is a vectorless RAG SDK: it builds hierarchical document trees, then uses LLM-guided tools to search them and answer questions.

  50. google-research/google-research★ 38.9k · Jupyter Notebook

    Google Research is a collection of independently maintained research projects, not one integrated application; its README recommends working within a project subdirectory.

Counted over GitDiagram's stored diagrams of public repositories, from the names their components and groups carry and the files they link to; the rules are the ones in the 2026 open-source architecture report. Stars are GitHub's when the diagram was last made. Rebuilt daily: hubs grow as more repositories' languages are known.