II-Agent is an Apache-2.0 open-source Python framework for building, running, forking, and extending intelligent agents, aimed at solo developers, research teams, and enterprises that want full control over models and cost through bring-your-own-key (BYOK) configuration.

What it is

II-Agent is an agentic AI framework written in Python and published under the Apache-2.0 licence. It ships as a full stack rather than a library alone: a backend, a frontend, and supporting infrastructure that together run agentic workflows across chat, research, document, slide, and code generation tasks. The project positions itself as "out of beta," with 3,387 stars and 524 forks at the time of this listing, and it carries topic tags including agent, agentic-ai, ai-agent, ai-agent-framework, and llm. Development is active, with the most recent push dated 2026-08-16 and 15 open issues.

The concrete problem it solves is vendor lock-in and opacity in hosted agent platforms. The README states the project is "100% open source," with "no black boxes" and "no vendor lock-in," and it lets operators supply their own API keys for full control over cost and model providers. It replaces the closed, single-provider agent product by giving teams the same agentic surface — planning, tool use, document generation, app integrations — as software they can run, fork, and modify themselves. It lives in the broader LLM agent ecosystem, alongside other agentic frameworks, but distinguishes itself by bundling a working application stack rather than only orchestration primitives.

Key capabilities

• Build workflows covering mobile app development, website app development, and storybook generation from a single prompt, plus video and image generation with multiple models inside one workflow.
• Live Editing for real-time editing of websites, slides, and storybooks, and Plan Mode for visual project planning before generation begins.
• Fast Research and Deep Research modes, the latter for multi-step investigations, with interactive website generation that turns research briefs into sites with structure, visuals, citations, and embedded Q&A.
• Built-in and custom Skills for reusing workflows and connecting GitHub-based processes, plus App Integrations with Gmail, Slack, GitHub, Notion, Google Calendar, Discord, Dropbox, and Canva.
• Chat with multi-model conversations that can switch providers mid-thread, file attachments, a code interpreter, and text file search.
• Documents support for PDF extraction and creation, Excel formulas and charts, Word editing, and PowerPoint manipulation, alongside Slides prompt-to-deck creation with templates and live collaborative editing.
• Configuration through MODEL_CONFIGS inline JSON in .env or a model_configs.yaml file referenced by MODEL_CONFIGS_FILE, with model_configs.example.yaml as the starting point.

Who uses it and how

• Solo developers who want an agent stack they can run locally and extend without a platform account, selecting a default model such as a Claude Sonnet entry via is_default in the model config.
• Research teams running fast or deep investigations and publishing the output as structured websites with citations and embedded Q&A.
• Enterprises building internal tooling, forking the codebase and integrating it with their own Gmail, Slack, GitHub, Notion, or Google Calendar accounts through the App Integrations layer.
• Teams that need document and deck automation — PDF, Excel, Word, PowerPoint, and Slides — as an agent-driven workflow rather than manual authoring.
• Operators who prefer a hosted option first, using the web app at https://agent.ii.inc/ and the Discord community for support before committing to a self-hosted deployment.

Getting started

Clone the repository and run make setup to create the .env files and install dependencies, then configure at least one LLM provider key, and run make dev-all to start the backend, frontend, and infrastructure together. The prerequisites are Docker, uv, and Node.js with npm, and the local stack brings up the backend at http://localhost:8000, the frontend at http://localhost:1420, PostgreSQL at localhost:5432, Redis at localhost:6379, and MinIO at http://localhost:9001.

How it compares

The facts supplied here name no paid products that II-Agent replaces, and the README does not name competing frameworks either; the manus topic tag only signals the category of hosted agentic products it sits near. It therefore stands alone in this registry rather than being contrasted against specific alternatives.

When to use it — and when not to

Running II-Agent means operating real infrastructure: Docker, PostgreSQL, Redis, and MinIO as S3-compatible storage, plus .env configuration for the database, storage, auth, and LLM keys, and a frontend .env for the API URL, Google OAuth, and theme. Anyone who does not want to run a database, a cache, and object storage, or who does not want to manage their own model keys, should use the hosted web app instead or pick a managed product. Note also that this listing's README excerpt ends mid-section at the model configuration table, so prospective self-hosters should check the repository for the complete configuration and deployment documentation before planning a production rollout.

# 1. Clone the repository
git clone https://github.com/Intelligent-Internet/ii-agent.git
cd ii-agent

# 2. Run first-time setup (creates .env files + installs deps)
make setup

# 3. Configure your LLM API keys
#    Edit .env and set at least one LLM provider:

#    Option A: Inline JSON in .env
#    MODEL_CONFIGS='[{"model_id":"claude-sonnet-4-6","provider":"Anthropic","api_key":"sk-ant-...","display_name":"Claude Sonnet 4","is_default":true}]'

#    Option B: YAML config file
#    Copy model_configs.example.yaml to model_configs.yaml, fill in your keys,
#    then set MODEL_CONFIGS_FILE=model_configs.yaml in .env

# 4. Start everything (infra + backend + frontend)
make dev-all

make help             # Show all available commands
make dev-all          # Start everything (infra + backend + frontend)
make infra            # Start only Postgres, Redis, MinIO
make backend-dev      # Start backend only (port 8000)
make frontend-dev     # Start frontend only (port 5173)
make db-migrate       # Run database migrations
make lint             # Lint backend + frontend
make format           # Auto-format backend + frontend
make test             # Run all tests
make stack            # Start full stack via Docker Compose

# Copy and edit the stack env file
cp docker/.stack.env.example docker/.stack.env
# Edit docker/.stack.env with your credentials

make stack            # Start full stack
make stack-build      # Start with --build (rebuild images)
make stack-down       # Stop and clean up
make stack-logs       # Tail all logs