RESTai tackles a challenge many AI platforms face: how to unify multi-project AI capabilities and expose them through a single, manageable interface. It combines several AI techniques — retrieval-augmented generation (RAG) layered with a knowledge graph, autonomous agents, inference pipelines, and visual logic — under one roof, accessible via a REST API and paired with a built-in React dashboard.
what RESTai offers and how it is built
At its core, RESTai is an AI-as-a-Service (AIaaS) platform designed to support diverse AI workloads across projects with a unified RESTful interface. It supports multiple LLM providers including OpenAI, Anthropic, Ollama, Gemini, LiteLLM, vLLM, and Microsoft Azure, providing flexibility to pick or switch underlying models without rewriting client code.
The architecture incorporates several advanced AI concepts:
RAG with knowledge graph: RESTai doesn’t just perform retrieval-augmented generation; it enhances it by building a knowledge graph on top of extracted named entities (NER). This means the retrieval context is enriched and structured, potentially improving relevance and inference quality.
Agentic browser per chat session: This is one of the most technically interesting aspects. Each chat runs isolated in its own Docker container that hosts a Playwright-powered browser instance. Secrets are encrypted and never exposed in plaintext to the language model context, while domain allowlists prevent prompt injection attacks. Persistent sessions are managed in Redis with a 30-day TTL, enabling continuity.
MCP server integration: RESTai integrates with the MCP server to extend agent capabilities, allowing agents to use external tools and services securely.
Blockly visual logic builder: For users who want to compose AI pipelines without writing code, RESTai offers a Blockly-based visual editor. This lowers the barrier for non-developers or rapid prototyping.
The platform’s backend is Python-based, distributed as a PyPI package (restai-core) with a pre-built React frontend bundled — so no Node.js build step is required. There’s also a multi-architecture Docker image supporting both linux/amd64 and linux/arm64, making deployment on diverse hardware easier.
the agentic browser architecture: isolation and security tradeoffs
The agentic browser design is where RESTai stands out. Opening a real browser controlled programmatically for each chat session is non-trivial:
Isolation: Running Playwright in per-chat Docker containers ensures that each session is sandboxed from others. This containment is crucial to avoid cross-session data leakage or interference.
Secrets management: RESTai encrypts secrets at rest and ensures they never appear in plaintext inside the LLM context. This is a strong security measure, preventing accidental exposure through prompt history or logs.
Domain allowlists: To prevent prompt injection or malicious browsing, RESTai restricts browser navigation to approved domains per project. This is a practical tradeoff balancing flexibility versus security.
Session persistence: Using Redis to store session state with a 30-day TTL means chats can continue over time without loss of context — important for practical applications.
This architecture is production-ready but comes with operational considerations. Running one Docker container per chat, each with a browser, can be resource-intensive. Scaling horizontally will require robust container orchestration, monitoring, and cost awareness.
quick start with RESTai
Getting started is straightforward thanks to clear installation instructions and multiple deployment options:
Install from PyPI
pip install restai-core
restai init # Create database + admin user
restai migrate # Run migrations
restai serve # → http://localhost:9000/admin (admin / admin)
You can use an environment file for configuration, for example:
restai serve -e .env -p 8080 -w 4
The PyPI package bundles the pre-built React dashboard, so there’s no need for a separate Node.js build step.
Run from source (development)
git clone https://github.com/apocas/restai && cd restai
make install
make dev # → http://localhost:9000/admin (admin / admin)
Docker
The project provides multi-architecture prebuilt images that you can run directly:
docker run -p 9000:9000 apocas/restai:latest
You can also pass environment variables and mount volumes for persistent data:
docker run -d --name restai -p 9000:9000 --env-file .env \
-v restai-data:/app/data \
apocas/restai:latest
This flexibility covers a range of deployment preferences from local dev to containerized production.
verdict: who should consider RESTai
RESTai is a solid pick if you need a unified AIaaS platform that combines RAG, agentic browsing, multi-provider LLM support, and visual AI pipeline composition. Its architecture reflects a mature approach to security and session management, especially around the agentic browser integration.
The operational tradeoff is the resource cost and complexity of running isolated browser containers per chat session. This could limit scalability without investment in container orchestration and monitoring.
For teams building AI products that require real browser automation with strong isolation, or those who want a ready-made platform to experiment with multi-model AI pipelines and agents, RESTai offers a compelling, well-engineered base.
The code quality appears clean, and the developer experience is improved by bundling the frontend and clear quickstart commands. However, this platform may be overkill for straightforward AI API needs without the agentic browser or RAG knowledge graph demands.
Overall, RESTai is worth exploring for AI practitioners who want a flexible, extensible AI platform with built-in security patterns and operational maturity around browser-based agents.
Related Articles
- LocalAI: running diverse AI models locally with multi-backend support and agent capabilities — LocalAI enables running 36+ AI models locally without GPU, supporting multi-user API access and built-in AI agents with
- AgentGPT: building autonomous AI agents with a full-stack web platform — AgentGPT offers a full-stack solution to deploy autonomous AI agents in the browser using Next.js, FastAPI, and Langchai
- CrewAI: A lean Python framework for orchestrating autonomous AI agents with precise control — CrewAI is a Python framework for autonomous AI agents emphasizing speed, flexibility, and precise control through ‘Crews
- Building a production-ready second brain with agentic RAG and LLMOps — Explore an open-source course that teaches building a production-grade AI assistant using advanced retrieval-augmented g
- Inside AI Engineering Hub: a hands-on collection of production-ready AI projects — AI Engineering Hub offers 90+ production-ready AI projects spanning LLMs, RAG, AI agents, and MCP, organized by difficul
→ GitHub Repo: apocas/restai ⭐ 502 · Python