Project Settings

Once a project is actively used inside Origin, teams need control over how agents execute work, not just what they can access. The Settings tab defines these execution-level behaviors. It governs the degree of autonomy agents are granted, how changes flow into review workflows, and how context is retained across sessions.

These settings do not expand repository access or override permission boundaries. Instead, they shape how agent-driven work is carried out within the project’s existing security and access constraints.

Developers typically configure this tab when transitioning a project from initial exploration to sustained development, or when aligning agent behavior with team review practices and iteration speed.

Autopilot Mode

Autopilot Mode controls whether agents can proceed with specific safe actions without requiring step-by-step user approval.

When enabled:

• agents can execute non-destructive, reversible actions autonomously,
• operations that modify code structure, apply diffs, or affect project state still respect safety guardrails,
• high-impact or destructive actions continue to require explicit confirmation.

Autopilot does not grant additional capabilities. It only removes manual confirmation for actions that Origin already classifies as low risk.

This setting is commonly enabled in projects where agents perform repetitive or well-understood work, such as formatting, incremental refactoring, test adjustments, or analysis tasks. It allows developers to iterate faster without reviewing every intermediate step, while still retaining control over meaningful changes.

Auto-PR

Auto-PR controls how code changes produced by agents are surfaced for review.

When enabled:

• agent executions that result in code changes automatically create a pull request,
• changes are grouped and attributed to the task that generated them,
• existing GitHub review and approval workflows remain unchanged.

This setting is useful in projects where tasks represent discrete units of work and teams want agent output to integrate directly into their normal code review process. Instead of manually extracting diffs from sessions, developers can review agent-generated changes as standard pull requests, with clear traceability back to the originating task.

Memories

Memories control whether agents retain conversational and decision context across sessions within the same project.

When enabled:

• agents remember prior discussions, clarifications, and decisions,
• long-running work maintains continuity across chats,
• repeated explanations and restated assumptions are avoided.

When disabled:

• each session starts without a conversational history,
• agents rely only on repository content and explicitly provided context.

Teams often enable Memories when working on multi-day initiatives, evolving architecture, or iterative task refinement. Some teams intentionally disable it when validating behavior from a clean baseline or experimenting with alternative approaches without influence from earlier sessions.

Wiki

The Wiki setting enables project-level documentation authored directly inside Origin.

When enabled:

• teams can create lightweight internal documentation pages,
• documentation becomes part of the project’s shared context,
• agents can reference this material during reasoning and task execution.

The Wiki does not replace repository documentation. It is typically used for information that is contextual, evolving, or not suitable for version control, such as architectural notes, onboarding guidance, or internal conventions that agents should consistently reference.

Semantic Search

Semantic Search improves how agents retrieve and relate information across the project’s context.

When enabled:

• agents retrieve information based on meaning rather than keywords,
• relevant documents and repository sections surface more reliably,
• reasoning across loosely related files and documents improves.

This setting affects how context is retrieved, not what data is stored or accessed. It is most useful in projects with large documentation sets or accumulated context where keyword matching becomes insufficient.

How These Settings Interact

The Settings tab defines execution behavior, not access control.

In practice:

• Autopilot and Auto-PR determine how automated execution flows into review,
• Memories and Semantic Search determine how context is retained and retrieved,
• Wiki determines whether internal documentation participates in agent reasoning.

Teams typically revisit these settings as confidence in agent workflows grows, adjusting autonomy and memory as projects move from experimentation to sustained development.