diff --git a/docs/.gitbook/assets/atomic-agent-logo.png b/docs/.gitbook/assets/atomic-agent-logo.png new file mode 100644 index 000000000..18ff3a7ce Binary files /dev/null and b/docs/.gitbook/assets/atomic-agent-logo.png differ diff --git a/docs/SUMMARY.md b/docs/SUMMARY.md index 80188f277..ad4a3f4dc 100644 --- a/docs/SUMMARY.md +++ b/docs/SUMMARY.md @@ -765,6 +765,7 @@ * [Our Integration List](integrations/our-integration-list.md) * [Agno](integrations/agno.md) * [Aider](integrations/aider.md) +* [⚛️ Atomic Agent](integrations/atomic-agent.md) * [AutoGPT](integrations/autogpt.md) * [Browser Use](integrations/browser-use.md) * [Cline](integrations/cline.md) diff --git a/docs/integrations/atomic-agent.md b/docs/integrations/atomic-agent.md new file mode 100644 index 000000000..c193b073e --- /dev/null +++ b/docs/integrations/atomic-agent.md @@ -0,0 +1,167 @@ +# ⚛️ Atomic Agent + +
Atomic Agent logo
+ +## About + +Atomic Agent is an open-source, local-first AI agent that runs on your computer. It plans multi-step tasks and executes them with real tools: shell, files, browser automation, skills, and MCP servers. All agent state, memory, and configuration live on your machine in `~/.atomic-agent`. + +Atomic Agent can run open models fully locally through a built-in llama.cpp backend, or connect to cloud models. AI/ML API is a **native built-in provider**: one API key gives the agent access to 500+ models, including GPT, Claude, Gemini, Qwen, and Grok families, with no endpoint configuration. + +Atomic Agent is MIT-licensed and available for macOS, Linux, and Windows. + +{% hint style="success" %} +**No native Anthropic or Google provider exists in Atomic Agent.** AI/ML API is the built-in way to use Claude and Gemini models with it. +{% endhint %} + +### What you get + +* A terminal agent (CLI + TUI) that plans and executes multi-step tasks +* Tool use out of the box: shell, file operations, browser automation, skills, MCP servers +* Local-first state: config, memory, and sessions stored in `~/.atomic-agent` +* Mix of local and cloud models in one app (for example, local embeddings + cloud chat) +* Live model catalog: the provider fetches the current AI/ML API model list at setup time +* Telegram connector and an OpenAI-compatible local HTTP server +* MIT license, installs without an account + +*** + +## Quick Install + +One command, no account or key required: + +```sh +# macOS / Linux +curl -fsSL https://atomicagent.io/install | sh +``` + +```powershell +# Windows +irm https://atomicagent.io/install.ps1 | iex +``` + +*** + +## Configuration + +You will need an AIMLAPI key from your [account dashboard](https://aimlapi.com/app/keys). + +### Option 1: Built-in provider wizard (recommended) + +Atomic Agent ships with a native AI/ML API provider, so there is nothing to configure by hand. + +1. Run `atomic-agent tui`. +2. Type `/model` to open the LLM Local/Cloud panel. +3. Select **AI/ML API (aimlapi.com — 500+ models, OpenAI-compatible)**. +4. Paste your API key when prompted. +5. Pick a chat model. The list is fetched live from the AI/ML API, so new models appear as soon as they are released. + +The wizard stores the key as `AIMLAPI_API_KEY` in `~/.atomic-agent/.env` and writes the provider entry to `~/.atomic-agent/config.json`. + +### Option 2: Manual configuration + +Add the provider to the `llm` block of `~/.atomic-agent/config.json`: + +```json +{ + "llm": { + "activeTextProvider": "aimlapi", + "activeEmbeddingProvider": "local-llama", + "toolTransport": "auto", + "providers": [ + { + "id": "aimlapi", + "kind": "aimlapi", + "defaultChatModel": "openai/gpt-5.5-2026-04-23" + } + ] + } +} +``` + +Then put your key in `~/.atomic-agent/.env`: + +```sh +AIMLAPI_API_KEY=your_key_here +``` + +The base URL (`https://api.aimlapi.com`) is built into the provider; you do not need to set it. + +{% hint style="info" %} +If you prefer the generic `"kind": "openai-compatible"` provider instead, set `"baseUrl": "https://api.aimlapi.com"` **without** the `/v1` suffix. Atomic Agent appends `/v1/chat/completions` itself, and a doubled path returns 404. That provider kind reads its key from `OPENAI_COMPAT_API_KEY`. +{% endhint %} + +*** + +## Model Selection + +* Default chat model for the AI/ML API provider: `openai/gpt-5.5-2026-04-23`. +* Any model ID from the [AI/ML API catalog](https://aimlapi.com/models) works, for example `anthropic/claude-sonnet-4-5` or `alibaba/qwen3.8-max-preview`. The configured model ID is sent to the API as-is, so you are not limited to the bundled list. +* Embedding models from the AI/ML API catalog can be selected too, or embeddings can stay on the local llama-server daemon. +* For cloud models, Atomic Agent uses native function calling, so tool use works out of the box. + +*** + +## Verification + +Start a session: + +```sh +atomic-agent tui +``` + +Ask for something that requires a tool call, for example: "list the files in this folder and summarize them". If the model responds and tools execute, the integration is working. + +*** + +## Use Cases + +* **Frontier models inside a local agent.** Keep the agent runtime, tools, and files on your machine while using top cloud models for reasoning. +* **Claude and Gemini in Atomic Agent.** These model families are only reachable through an aggregator; AI/ML API is the supported path. +* **One key, many models.** Switch between model families without separate accounts and billing per provider. +* **Mixed setups.** Run embeddings locally and chat in the cloud, or the other way around. + +*** + +## Troubleshooting + +
+ +401 / authentication errors + +Check that `AIMLAPI_API_KEY` is present in `~/.atomic-agent/.env` and has no extra whitespace. Re-entering the key via the `/model` wizard rewrites it correctly. + +
+ +
+ +404 on chat requests (manual openai-compatible setup) + +Remove the `/v1` suffix from `baseUrl`. The agent appends `/v1/chat/completions` itself. + +
+ +
+ +Model not found + +Verify the exact model ID against the [AI/ML API models list](https://aimlapi.com/models). The ID is passed to the API verbatim. + +
+ +
+ +Switching provider or model later + +Open `/model` in the TUI at any time to change the provider or the chat model. No restart is required. + +
+ +*** + +## Links + +* Website: [atomicagent.io](https://atomicagent.io) +* GitHub: [github.com/AtomicBot-ai/atomic-agent](https://github.com/AtomicBot-ai/atomic-agent) +* Discord: [discord.gg/Us7qXtDGw](https://discord.gg/Us7qXtDGw) +* X: [@atomicagent_io](https://x.com/atomicagent_io) diff --git a/docs/integrations/our-integration-list.md b/docs/integrations/our-integration-list.md index ad67c444a..e2ed05bd5 100644 --- a/docs/integrations/our-integration-list.md +++ b/docs/integrations/our-integration-list.md @@ -6,4 +6,4 @@ description: About third-party integrations Our API endpoint can be integrated with popular AI workflow platforms and tools, allowing their users to access our models through these environments. -
ServiceDescription
AgnoA lightweight library for building agents — AI programs that operate autonomously, use tools, and have memory, knowledge, storage, and reasoning capabilities.
AiderA command-line pair programming tool that connects to OpenAI-compatible APIs. It lets you chat with models to edit your codebase, auto-commit changes, and build software collaboratively from the terminal.
AutoGPTAn open-source platform designed to help you build, test, and run AI agents using a no-code visual interface. It allows users to link LLMs with tools, memory, planning modules, and action chains.
Browser UseAn LLM-powered browser automation tool that allows language models to control a real browser via structured system and user prompts. The LLM decides which browser actions to perform (open pages, navigate, click, read content), while the execution is handled by the browser runtime. The integration works with any supported LLM and does not rely on model-specific features.
Claude CodeA powerful coding agent by Anthropic that can analyze codebases, generate code, and assist with development workflows. By pointing it to AIML API, you can use Claude models available on our platform without connecting directly to Anthropic.
ClineAn open-source AI coding assistant with two working modes (Plan/Act), terminal command execution, and support for the Model Context Protocol (MCP) in VS Code.
continue.devAn open-source IDE extension and hub for rules, tools, and models that let you create, share, and use custom AI code assistants.
ElizaOSA powerful multi-agent simulation framework designed to create, deploy, and manage autonomous AI agents. Built with TypeScript, it provides a flexible and extensible platform for developing intelligent agents that can interact across multiple platforms while maintaining consistent personalities and knowledge.
GPT ResearcherAn autonomous agent that takes care of the tedious task of research for you, by scraping, filtering and aggregating over 20+ web sources per a single research task.
HermesAn agent framework for CLI and messaging workflows that supports OpenAI-compatible providers. It can connect to AI/ML API through a native provider fork or a manual custom backend setup.
Hermes DesktopA native desktop app for installing, configuring, and chatting with Hermes Agent. Supports AI/ML API as a built-in provider preset for easy access to OpenAI-compatible models through a GUI.
Kilo CodeAn open-source AI coding assistant and VS Code extension that enables natural-language code generation, debugging, and refactoring through customizable modes (Architect, Code, Debug, etc.). It supports multiple model providers, integrates with the Model Context Protocol (MCP), and allows developers to extend functionality with custom tools and workflows.
LangflowA new visual framework for building multi-agent and RAG applications. It is open-source, Python-powered, fully customizable, and LLM and vector store agnostic. Its intuitive interface allows for easy manipulation of AI building blocks, enabling developers to quickly prototype and turn their ideas into powerful, real-world solutions.
LiteLLMAn open-source Python library that provides a unified API for interacting with multiple large language model providers. It allows developers to switch between different models with minimal code changes, optimizing cost and performance. LiteLLM simplifies integration by offering a single interface for various LLM endpoints, enabling seamless experimentation and deployment across different AI providers.
MakeA powerful, enterprise-scale automation platform. It offers flow control, data manipulation, HTTP/webhooks, AI agents and tools, notes, an MCP server, and many other features at your service.
ManusA workflow and AI-agent orchestration platform that lets users integrate custom APIs, define automation logic, and run LLM-powered tools inside a unified interface. Manus supports custom model backends (such as AI/ML API), prompt templates, request routing, secure secret storage, and visual debugging.
MarvinA Python framework by PrefectHQ for building agentic AI workflows and producing structured outputs. It allows developers to define Tasks (objective-focused units of work) and assign them to specialized Agents (LLM-powered configurations). Marvin supports type-safe results via Pydantic models, integrates with multiple LLM providers through Pydantic AI, and enables orchestration of multi-agent threads for complex workflows.
Money Printer TurboA one-click short video generator powered by AI LLMs. Uses AI/ML API as a built-in provider for script and story generation — configure via config.toml or the built-in WebUI.
n8nAn open-source workflow automation tool that lets you connect various services and automate tasks without writing full integrations manually.
Oh My PiAn AI coding agent for the terminal with subagents, plan mode, LSP, and DAP. Supports AI/ML API as a built-in provider preset — set AIMLAPI_API_KEY and omp discovers 350+ chat-compatible models automatically.
OpenClaw

A personal AI assistant that runs on your own devices and connects to popular messaging platforms (such as WhatsApp, Telegram, Slack, Discord, and others) while preserving full data privacy (all agent data is stored locally in a SQLite database).

Developers use OpenClaw to build multi-channel AI assistants with streaming responses, browser automation, vision, and voice features. It includes a local Gateway service, a CLI for management, and support for 12+ messaging platforms.

Roo CodeAn autonomous AI programming agent that works right inside your editor, such as VS Code. It helps you code faster and smarter — whether you're starting a new project, maintaining existing code, or exploring new technologies.
SillyTavernA locally installed user interface that allows you to interact with text generation LLMs, image generation engines, and TTS voice models. Integration with the AI/ML API currently applies only to LLMs.
ToolhouseA Backend-as-a-Service (BaaS) to build, run, and manage AI agents. Toolhouse simplifies the process of building agents in a local environment and running them in production.
+
ServiceDescription
AgnoA lightweight library for building agents — AI programs that operate autonomously, use tools, and have memory, knowledge, storage, and reasoning capabilities.
AiderA command-line pair programming tool that connects to OpenAI-compatible APIs. It lets you chat with models to edit your codebase, auto-commit changes, and build software collaboratively from the terminal.
Atomic AgentAn open-source, local-first AI agent for the terminal that plans multi-step tasks and executes them with real tools: shell, files, browser automation, skills, and MCP servers. Ships with AI/ML API as a native built-in provider — pick it in the setup wizard, paste your key, and choose any of 500+ models fetched live from the catalog.
AutoGPTAn open-source platform designed to help you build, test, and run AI agents using a no-code visual interface. It allows users to link LLMs with tools, memory, planning modules, and action chains.
Browser UseAn LLM-powered browser automation tool that allows language models to control a real browser via structured system and user prompts. The LLM decides which browser actions to perform (open pages, navigate, click, read content), while the execution is handled by the browser runtime. The integration works with any supported LLM and does not rely on model-specific features.
Claude CodeA powerful coding agent by Anthropic that can analyze codebases, generate code, and assist with development workflows. By pointing it to AIML API, you can use Claude models available on our platform without connecting directly to Anthropic.
ClineAn open-source AI coding assistant with two working modes (Plan/Act), terminal command execution, and support for the Model Context Protocol (MCP) in VS Code.
continue.devAn open-source IDE extension and hub for rules, tools, and models that let you create, share, and use custom AI code assistants.
ElizaOSA powerful multi-agent simulation framework designed to create, deploy, and manage autonomous AI agents. Built with TypeScript, it provides a flexible and extensible platform for developing intelligent agents that can interact across multiple platforms while maintaining consistent personalities and knowledge.
GPT ResearcherAn autonomous agent that takes care of the tedious task of research for you, by scraping, filtering and aggregating over 20+ web sources per a single research task.
HermesAn agent framework for CLI and messaging workflows that supports OpenAI-compatible providers. It can connect to AI/ML API through a native provider fork or a manual custom backend setup.
Hermes DesktopA native desktop app for installing, configuring, and chatting with Hermes Agent. Supports AI/ML API as a built-in provider preset for easy access to OpenAI-compatible models through a GUI.
Kilo CodeAn open-source AI coding assistant and VS Code extension that enables natural-language code generation, debugging, and refactoring through customizable modes (Architect, Code, Debug, etc.). It supports multiple model providers, integrates with the Model Context Protocol (MCP), and allows developers to extend functionality with custom tools and workflows.
LangflowA new visual framework for building multi-agent and RAG applications. It is open-source, Python-powered, fully customizable, and LLM and vector store agnostic. Its intuitive interface allows for easy manipulation of AI building blocks, enabling developers to quickly prototype and turn their ideas into powerful, real-world solutions.
LiteLLMAn open-source Python library that provides a unified API for interacting with multiple large language model providers. It allows developers to switch between different models with minimal code changes, optimizing cost and performance. LiteLLM simplifies integration by offering a single interface for various LLM endpoints, enabling seamless experimentation and deployment across different AI providers.
MakeA powerful, enterprise-scale automation platform. It offers flow control, data manipulation, HTTP/webhooks, AI agents and tools, notes, an MCP server, and many other features at your service.
ManusA workflow and AI-agent orchestration platform that lets users integrate custom APIs, define automation logic, and run LLM-powered tools inside a unified interface. Manus supports custom model backends (such as AI/ML API), prompt templates, request routing, secure secret storage, and visual debugging.
MarvinA Python framework by PrefectHQ for building agentic AI workflows and producing structured outputs. It allows developers to define Tasks (objective-focused units of work) and assign them to specialized Agents (LLM-powered configurations). Marvin supports type-safe results via Pydantic models, integrates with multiple LLM providers through Pydantic AI, and enables orchestration of multi-agent threads for complex workflows.
Money Printer TurboA one-click short video generator powered by AI LLMs. Uses AI/ML API as a built-in provider for script and story generation — configure via config.toml or the built-in WebUI.
n8nAn open-source workflow automation tool that lets you connect various services and automate tasks without writing full integrations manually.
Oh My PiAn AI coding agent for the terminal with subagents, plan mode, LSP, and DAP. Supports AI/ML API as a built-in provider preset — set AIMLAPI_API_KEY and omp discovers 350+ chat-compatible models automatically.
OpenClaw

A personal AI assistant that runs on your own devices and connects to popular messaging platforms (such as WhatsApp, Telegram, Slack, Discord, and others) while preserving full data privacy (all agent data is stored locally in a SQLite database).

Developers use OpenClaw to build multi-channel AI assistants with streaming responses, browser automation, vision, and voice features. It includes a local Gateway service, a CLI for management, and support for 12+ messaging platforms.

Roo CodeAn autonomous AI programming agent that works right inside your editor, such as VS Code. It helps you code faster and smarter — whether you're starting a new project, maintaining existing code, or exploring new technologies.
SillyTavernA locally installed user interface that allows you to interact with text generation LLMs, image generation engines, and TTS voice models. Integration with the AI/ML API currently applies only to LLMs.
ToolhouseA Backend-as-a-Service (BaaS) to build, run, and manage AI agents. Toolhouse simplifies the process of building agents in a local environment and running them in production.