Odysseus AI

Odysseus AI is a private AI workspace for running structured deep research, autonomous agent tasks, and memory workflows in an isolated environment. It is designed for users who need controlled, source-traceable AI workflows rather than one-off prompts or scattered local scripts.

Key Features

• Model Routing & BYOK: Managed routing through OpenRouter with optional BYOK for users who want control over billing and model choices, including support for DeepSeek, Llama, and open-weight models.
• Autonomous Agents & MCP: Deploy agents with secure, read-only access to selected files, memory, and custom MCP server context. Zero default access with explicit authorization per session.
• Deep Research: Turn URLs, PDFs, Markdown, TXT files, and notes into structured, source-traceable briefs with verifiable source ledgers.
• Isolated Memory: Saved sources, notes, and prior reports remain available as reusable context for new agent runs, building a private knowledge base over time.
• Zero-Data-Retention (ZDR) Routing: Odysseus does not train models on user files, prompts, notes, sources, or outputs.

Use Cases

• Academic and technical literature review with structured citation-style source tables.
• Competitor and product benchmarking briefs with traceable evidence.
• Market intelligence and OSINT producing structured, reviewable market briefs.
• Due diligence and investment research with claims traceable back to underlying materials.

Workflow

1. Connect — Securely upload PDFs, Markdown, TXT files, and notes into an isolated workspace.
2. Structure — The engine organizes chosen sources into a structured context window.
3. Delegate — Assign research parameters to an autonomous agent with strict execution limits.
4. Review — Inspect the generated brief, verify the source table, and audit agent actions.
5. Continue — Retain insights in isolated memory for follow-up tasks.

Key Benefits

• Private by design with no model training on user data
• Zero-Data-Retention (ZDR) routing where supported
• OpenRouter BYOK for full control over model choices and billing
• Isolated workspace memory retains context across workflows