Illospace

Product Introduction

1. Definition: Illospace is an open-source, full-stack collaborative AI workspace and agentic runtime environment. Categorized as a Multi-Agent Orchestration Platform (MAOP), it integrates a FastAPI backend with a SvelteKit frontend to provide a unified "team brain" where human users and autonomous AI agents interact within a shared execution context.

2. Core Value Proposition: Illospace exists to eliminate the siloed nature of traditional AI tools by providing a persistent, shared memory layer and a collaborative canvas for human-agent teamwork. By utilizing vector-based long-term memory (pgvector) and "Cortex" thought threads, it enables agents to maintain state, access system-level tools, and execute background tasks like cron jobs and database management, ensuring that AI work is transparent, inspectable, and deeply integrated into the team's operational workflow.

Main Features

1. Shared Team Brain & Persistent Vector Memory: Illospace utilizes a PostgreSQL database equipped with the pgvector extension to create a durable "team brain." This architecture allows for semantic retrieval of historical context, project data, and previous agent interactions. Unlike transient chat interfaces, this memory system ensures that agents and humans have access to the same knowledge base, enabling high-continuity workflows and reducing context drift in long-running projects.

2. Cortex Thought Threads & Agent Runtime: The platform features a specialized agent runtime that supports "Cortex" threads—complex execution paths where agents can deliberate, use tools, and interact with one another. This runtime manages model invocation, skill bundles, and browser-based tool execution. It allows agents to perform autonomous "runs" that are visible to the entire team, making the agent’s internal reasoning process (Chain of Thought) accessible and auditable via the SvelteKit dashboard.

3. Integrated System Tools & Vault Security: Illospace acts as a localized operating system for AI agents, offering built-in capabilities to create and manage databases, schedule recurring cron jobs, and develop workspace-specific applications. Security is handled through a vault-backed secret management system, allowing agents to securely access third-party API keys and credentials without exposing them to the frontend or unauthorized users, facilitating safe, high-privilege agent operations.

Problems Solved

1. Pain Point: Fragmented AI Context and "Black Box" Operations. Traditional AI assistants often operate in isolation, leading to context loss when switching between team members or tasks. Furthermore, the reasoning steps of autonomous agents are often hidden from users. Illospace addresses this by providing a unified canvas where all agent activities, thoughts, and tool executions are logged and visible in real-time.

2. Target Audience: This platform is designed for AI Engineers and Developers building agentic workflows, DevSecOps teams requiring automated system management, and high-velocity product teams who need to integrate LLM-based automation into their daily collaborative cycles. It is particularly valuable for organizations seeking a self-hosted, open-source alternative to proprietary AI collaboration platforms.

3. Use Cases: Essential for scenarios such as automated DevOps monitoring where agents manage cron jobs for system health; collaborative software development where agents and humans share a coding context; and dynamic data analysis where agents autonomously query databases and generate real-time interface visualizations for human review.

Unique Advantages

1. Differentiation: Unlike simple wrapper applications or chat-based LLM interfaces, Illospace is a full-stack infrastructure. It prioritizes "always-on" agents that can work independently of a live user session. While competitors often focus on single-user productivity, Illospace focuses on the multi-player aspect, treating the agent as a first-class team member with its own permissions, tools, and memory.

2. Key Innovation: The primary innovation lies in the tight coupling of a vector-enabled backend (FastAPI/pgvector) with a dynamic canvas-based frontend (SvelteKit). This "Canvas of Ideas" allows for the generation of dynamic interfaces, where agents don't just output text but can build and manipulate the UI components and workspace tools the team uses, effectively allowing agents to create their own operational dashboards.

Frequently Asked Questions (FAQ)

1. What are the system requirements for hosting Illospace locally? To run Illospace, you need Python 3.11+ and Node.js 22+. The backend requires a PostgreSQL 16+ instance with the pgvector extension installed. While it can run via Docker or Podman for ease of setup, a local "launcher" script (./illo) is provided to manage the environment, including an optional local embedding/GPU worker for lower-latency semantic searches.

2. Does Illospace support multiple LLM providers? Yes, Illospace is provider-agnostic. It supports major LLM providers such as OpenAI (including OAuth-based Codex sign-in) and can be configured to work with local models. The platform uses a flexible "provider adapter" architecture within the brain/platform directory, allowing users to swap or combine different inference engines for different agent skills.

3. Is Illospace suitable for production environments? Currently, Illospace is in an early open-source preview stage. While it includes robust features like vault-backed secrets and a structured FastAPI backend, the developers indicate that the API and project structure are actively evolving. For production-style self-hosting, users are encouraged to use the ./illo start command, which builds the static frontend and prepares the production API, but careful monitoring of the "early/preview" status is recommended.