Ódinn Forge docs

Ódinn Forge is a local-first personal AI workspace for conversations, projects, memory, tools, scheduled work, and the activity that connects them. Run it on a machine you control, then choose the model provider or local model server that fits your work.

Use the hubs below to get somewhere useful quickly.

Browse docs

Start hereInstall Ódinn, finish onboarding, and have your first conversation.

Use ÓdinnLearn the console, memory, tasks, projects, sessions, and scheduled work.

ConfigureConnect providers, remote hosts, backups, MCP extensions, and controls.

SecurityUnderstand the security model, privacy boundaries, isolation, and limits.

LabsExplore experimental capabilities with their own docs and boundaries.

ReferenceFind the CLI, API, schemas, state layout, and event ledger references.

HelpTroubleshoot common problems and collect useful diagnostics.

ProjectRead about the architecture, contribution workflow, and project history.

Start here

Install ÓdinnPut the Forge on your own machine and start the local gateway.

OnboardingWalk through the first-run setup and connect a model provider.

Your first conversationMeet the console and learn the smallest useful workflow.

Learn more

The operator consoleSee where conversations, context, tools, and activity come together.

Memory and contextUnderstand what Ódinn remembers and how to keep it useful.

Models and providersChoose, configure, and reason about the model behind your workspace.

PrivacyKnow what stays local and what a connected provider can receive.

ArchitectureFollow the main runtime pieces without starting in implementation details.

CLI referenceKeep the command surface close at hand when you need precision.

A useful mental model

Ódinn keeps the workspace, durable records, and local gateway together. The model is a replaceable collaborator; tools and scheduled work are explicit capabilities; the activity history gives you a readable account of what happened. Start with the workflow you want, then open the configuration and reference pages when you need more control.

There is no built-in product telemetry. A cloud model provider can still receive the prompts, recalled context, and tool results sent to its API, so read Privacy before connecting accounts or providers.