II Agent
A self-hosted, multi-model intelligent agent for building, research, and document work with your own API keys.
- Source repo
- Intelligent-Internet/ii-agent
- Stars
- ★ 3.4k
- Last updated
- 1mo ago
- License
- Apache-2.0
- Primary language
- Python
- FA score
- 28/100 · Major gaps
At a glance
- Works with
- Universal · cross-platformOpenAI API · Claude API
- You'll need
- Typical use
- A solo developer wants to start a mobile app or website from a short prompt while running the full frontend and backend stack locally.
- Main limitation
- Local operation requires Docker, uv, Node.js, and npm, plus PostgreSQL, Redis, and MinIO services.
- Source review
- 28/100 · Major gaps 4 safety controls not found
What does this agent do, and when should you use it?
II Agent is an Apache-2.0 open-source AI agent that individuals, research teams, and enterprises can run, fork, and extend. It uses a browser frontend and backend service with a local stack comprising PostgreSQL, Redis, and MinIO. Documented capabilities include prompt-driven mobile apps, websites, picture books, and slide decks, plus fast and multi-step research. It also covers multi-model chat, file attachments, a code interpreter, text-file search, and PDF, Excel, Word, and PowerPoint tasks. In the documented full development stack, the backend runs at localhost:8000 and the frontend runs at localhost:1420.
Run make setup to create .env files and install dependencies, then configure models either through MODEL_CONFIGS in .env or by copying model_configs.example.yaml to model_configs.yaml and setting MODEL_CONFIGS_FILE. make dev-all starts the infrastructure, backend, and frontend; the documented services are PostgreSQL, Redis, MinIO, the Backend at http://localhost:8000, and the Frontend at http://localhost:1420. In the running product, users can hold multi-model conversations, switch providers mid-thread, attach files, use a code interpreter and text-file search, and run fast or multi-step research that can turn a research brief into a website with structure, visuals, citations, and embedded Q&A. The README also lists prompt-driven mobile apps, websites, picture books, video and image generation, plus PDF extraction and creation, Excel formulas and charts, Word editing, and PowerPoint manipulation.
- A solo developer wants to start a mobile app or website from a short prompt while running the full frontend and backend stack locally.
- A research team needs to choose between fast and multi-step research, then deliver a research brief as an interactive website with citations and embedded Q&A.
- An internal-tools team wants to supply its own model API keys and configure OpenAI, Anthropic, or Google models in one system.
- A content creator wants to produce an illustrated picture book from one prompt or use multiple video and image models in one workflow.
- A knowledge-work user needs PDF extraction or creation, Excel formulas and charts, Word editing, or PowerPoint manipulation.
How do you install or deploy this agent?
Prerequisites: Docker, uv, Node.js, and npm.
- git clone https://github.com/Intelligent-Internet/ii-agent.git
- cd ii-agent
- make setup
- Set MODEL_CONFIGS for at least one model provider in .env, or copy model_configs.example.yaml to model_configs.yaml, add API keys, and set MODEL_CONFIGS_FILE=model_configs.yaml in .env.
- Run make dev-all.
How do you use this agent?
After startup, open http://localhost:1420; the backend is available at http://localhost:8000. Once at least one provider and API key are configured, use chat with file attachments, the code interpreter, or text-file search, and initiate build, research, or document tasks. To run only part of the stack, use make infra, make backend-dev, or make frontend-dev. For the containerized full stack, create and fill docker/.stack.env, then run make stack.
What are this agent's strengths and limitations?
- Bring-your-own-key configuration lets operators control model providers and costs, with documented OpenAI, Anthropic, and Google model configuration.
- It provides both a local development stack and a Docker Compose full-stack path with PostgreSQL, Redis, and S3-compatible MinIO.
- The documented product spans app generation, research-site generation, document handling, and multi-model chat instead of a single workflow.
- Local operation requires Docker, uv, Node.js, and npm, plus PostgreSQL, Redis, and MinIO services.
- At least one LLM provider API key is required; service availability and cost depend on the selected provider.
- The supplied material lists Gmail, Slack, GitHub, Notion, Google Calendar, Discord, Dropbox, and Canva integrations but does not provide authentication or setup steps for each one.
- The supplied material does not document a stable external API, SDK, or version-compatibility commitment.
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|
| II Agent This agent | 28 · Major gaps | ★ 3.4k | 1mo ago | Python | OpenAI API · Claude API |
| Loki Mode | 57 · Major gaps | ★ 1.1k | 2d ago | Shell | Claude Code · Claude API |
| Claude Code Skills & Agents | 50 · Major gaps | ★ 1.7k | 1mo ago | Python | Claude Code |
| MS-Agent | 48 · Major gaps | ★ 4.4k | 2d ago | Python | — |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence shows: The project uses BYOK model, where users bring their own API keys, which partially implements least privilege, but no fine-grained permission control or sandboxing is provided. User confirmation is absent; no documentation of user approval or review processes. Data flow transparency is missing; no explanation of how data is processed, stored, or transmitted. Sensitive data handling: .env files manage keys, but no encryption or protection measures are described. Dependency security: a large list of dependencies, but no evidence of vulnerability scanning or security audits. External effects: no documentation of potential external actions (e.g., sending emails, modifying files) and their risks. Rollback: no rollback mechanism provided. Source attribution: author information exists, but publisher is unverified. Deductions: lack of user confirmation, data flow transparency, external effects, and rollback mechanisms.
Evidence shows: Self-consistency: README and pyproject.toml descriptions are largely consistent, but code not deeply inspected. Dependency availability: dependency list is explicit, but availability not verified. Failure messages: no error handling or user-friendly failure prompts documented. Deductions: lack of failure message documentation.
Evidence shows: Audience and scenarios: README clearly targets developers, research teams, and enterprises, listing multiple use cases. Capability boundaries: feature list provided, but no explicit limits or boundaries. Trigger precision: no explanation of how to trigger agents or tasks. Environment fit: Docker and local installation options provided, supporting multiple OS. Deductions: lack of trigger precision and explicit capability boundaries.
Evidence shows: Information architecture: README is well-structured with table of contents and feature lists. Install notes: detailed installation steps and configuration instructions provided. Naming stability: project name and version are clear, but no API stability guarantee. Examples and FAQ: example configurations exist, but FAQ missing. Known limitations: none mentioned. License: Apache-2.0 clearly stated. Versioning and changelog: no CHANGELOG provided. Maintenance responsibility: author info exists, but no explicit maintenance commitment. Deductions: missing FAQ, known limitations, and changelog.
Evidence shows: Output usability: describes various output types (e.g., websites, apps, documents), but no output quality assurance. Marginal value: claims faster than earlier versions, but no benchmarks provided. Cost-benefit: BYOK model may reduce costs, but no cost analysis. Deductions: lack of concrete evidence.
Evidence shows: Claim traceability: feature claims in README are not linked to code or tests. Cross-source corroboration: no external validation. Fact-inference separation: no distinction between facts and inferences. Deductions: lack of verifiable evidence.
- Not found in source: confirmation before actingTurn on (or add) a confirmation step before it acts, and try it in a sandbox or test environment before real data.
- Not found in source: data-flow disclosureWatch which external services it contacts (proxy or firewall logs) and keep sensitive data out until you know where it goes.
- Not found in source: disclosed external effectsEstablish which external systems it writes to, sends to or changes, and verify with test accounts or repos before production.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- Publisher identity is unverified; proceed with caution.
- No user confirmation mechanism; the agent may perform actions without explicit user consent.
- Data flow is not transparent; review data handling practices yourself.
- Many dependencies; conduct your own security audit.
- No rollback mechanism; failures may be hard to recover from.