Memory that learns from you
Past conversations are distilled into compact facts and surfaced only when relevant.
Our agents remember your work, run code, browse the web, and work while you're away. Run models locally or reach for a frontier model like ChatGPT or Claude. Osaurus is always yours and stays local. Coming soon to Windows.

As seen in
Connect OpenAI, Anthropic, or Google Gemini with your own API key, or run open models locally. Choose a model for each agent and switch without losing its memory or skills.
For local models on Mac, Osaurus includes a native inference engine built for Apple Silicon.
Everything is stored on your Mac and signed at every boundary. Optional telemetry, no training on your data, no servers in the loop. When you do choose a cloud provider, the Privacy Filter scrubs names, emails, and secrets from the prompt first, shows you the redactions, and blocks the send rather than leak.
See exactly what is stored and what leaves →Your context, your memory, your tools, your identity. Other apps keep that layer on their servers. Osaurus keeps it on your computer, in the open, under your control.
Each agent has its own model, memory, and skills. Create specialists yourself, or let Osaurus build and coordinate them.
Point to a folder and agents can read, search, and edit its files. Shell, Python, and Node run in an isolated sandbox.
Agents browse sites, read pages, and fill forms. Each has its own persistent browser. Built-in web search needs no API keys.
Run agents on a schedule or when a folder changes. Automate daily journals, inbox summaries, and end-of-day commits.
The Osaurus app is free and MIT licensed. No subscription required. Use local models or connect a cloud provider. Cloud usage is billed separately.
Every team has a few people who know how everything works. Workspaces turns what they know into agents the whole team can ask, and the routines they carry into workflows that run on their own.
Osaurus for individuals stays free and MIT licensed. Workspaces is how teams pay for the parts a team needs.
The person who understands the codebase, contracts, or pricing builds an agent. The whole team asks it questions. Answers come from the source.
Agents call tools, read files, and run on schedules. Set up the release checklist, weekly report, or ticket triage once, and it keeps running for the whole team.
Shared agents run on your team's computers and talk over an encrypted network you host. Nobody stands up a server. Nobody edits a config file.
Use providers like Anthropic across your team, and choose a local model for agents handling sensitive data so that processing stays on your own hardware.
Your Research and Product Knowledge Assistants can use Claude, while your HR Assistant processes salary data with a local model. Choose what fits each agent’s work and your company’s data policies.
Workspaces is priced per seat, with a shared monthly pool of credits for cloud models. Local inference has no usage charges.
The IT lead who wants AI on hardware they control. The founder who wants the whole company moving faster without a tooling project. The business owner who would rather answer one question well than the same question ten times.
Osaurus for individuals stays free and MIT licensed. Workspaces adds paid features for sharing agents and workflows across your team.