Inference, engineered.

Arcanum is a local-first AI harness that brings your models, tools, files, and memory together. Give it a task, let it work through the steps, and keep control of the context it uses and the resources it spends. The Forge is the visual workspace being built around it.

Apply for Private Beta Discover Arcanum

Start with Arcanum. Help shape what comes next.

A native application: Native AOT builds for Apple Silicon Macs and Windows, with no separate .NET runtime to install.
Answers grounded in your material: Find relevant passages in workspace files and attachments, or explore the larger picture through layered summaries.
Work that follows through: Models and tools work through multiple steps, with delegation, progress checks, cancellation, and spending limits.
Memory with a source: Encrypted local records and OATH help useful context survive while keeping its origin and authority clear.

Your models. Your tools. One engine.

Keep the same work connected across your tools.

Your chosen local or hosted model providers

ArcanumModels, tools, and memory

Terminal and API clients

Memory governed by OATH

The Forge · Coming soon

Arcanum runs the work and keeps its records. Terminal clients and The Forge connect to the same engine.

Arcanum and The Forge

An engine for the work. A workspace to bring it together.

01 / Arcanum · Private beta

Arcanum

Give your AI the context to do useful work.

Connect the models you choose, work with your own files, and carry useful knowledge between conversations. Arcanum combines an agent engine with retrieval, durable memory, and controls you can inspect.

Work through tasks: Read, search, and edit workspace files with built-in and MCP tools. Delegate bounded jobs and use Eye of the World for a concise view of the workspace.
Find the right context: Use semantic search and RAG to find relevant passages. The Tapestry adds RAPTOR-style summary trees for larger collections; SQLite FTS5 helps find exact text.
Remember what matters: Lexicon stores structured facts, Saga recalls lessons from conversations, and Covenant separates context you confirm from suggestions the assistant makes.
Stay in control: Set budgets, configure Campaign tool policies with Sanctum, inspect tool activity, and manage backups and retention. Use Trials to check how your prompts perform.

Explore Arcanum and its memory system View Arcanum on GitHub

02 / The Forge · Coming soon

The Forge

A development environment for your AI work.

The Forge is a full-lifecycle Inference IDE powered by Arcanum: a place to write reusable instructions, compare results, follow agent work, and understand the context behind an answer.

Write: Edit prompts and reusable workflows with source and preview views.
Version: Save named revisions, compare changes, and choose which version to run.
Evaluate: Run Trials, compare outputs, and inspect traces of model and tool activity.
Follow the work: See conversations, agents, tools, and recorded spending in one workspace.

Give your AI work a place to grow.

Discover The Forge

Private Beta

Put Arcanum to work on something you care about.

Try it with your projects, tell us what helps, and help shape Arcanum and The Forge.

Apply for Private Beta