Nora | CLI-first agentic development workspace

Dark

Beta<br>$ cli-first$ local agents$ human approvals

Quick Start<br>macOSLinuxWindows

install.linux<br># install nora for linux<br>$curl -fsSL https://withnora.run/install.sh | bash

$ workspace preview$ session: local / mode: operator

$ supported clis

DownloadRead docs<br>$ available on<br>macOS<br>Windows<br>Linux

$ operating path<br>One path from intent to merge-ready output.<br>Scope

Scope<br>Execution boundaries and constraints

Spec

Spec<br>Requirements and acceptance criteria

Tasks

Tasks<br>Actionable units with ownership

Agents

Agents<br>Parallel implementation in worktrees

Diffs

Diffs<br>Review and validation checkpoints

PR

Open Pull Request<br>Approval-ready pull request

$ handoff<br>Hand tasks from one agent to another without losing context.<br>Keep the brief, current repo state, and task intent intact as work moves between agents instead of restarting the conversation every time a different tool takes over.

Flow<br>$ task: refine download flow<br>codex

claude

review

$ task to agent handoff<br>$ shared context<br>$ visible ownership

*:first-child]:order-2 lg:[&>*:last-child]:order-1 translate-y-10 opacity-0" style="transition-delay:70ms">$ scope<br>Derive tasks directly from scope.<br>Start from the shape of the work itself and let Nora break that into actionable units instead of manually rewriting the same plan for every agent session.

Planning<br>$ scope.txt<br>1. define installer paths<br>2. build download surface<br>3. wire release metadata

$ scope -> task breakdown<br>$ clear execution units<br>$ less prompt drift

$ skills<br>Attach the right skill set to the run.<br>Use a skill manager to bring specialized instructions and workflows into the session without bloating the default agent prompt.

Extensions<br>+ electron<br>+ frontend<br>+ tailwind<br>+ docs

$ skill manager<br>$ task-specific context<br>$ lighter defaults

*:first-child]:order-2 lg:[&>*:last-child]:order-1 translate-y-10 opacity-0" style="transition-delay:210ms">$ worktrees<br>Give every agent its own isolated worktree.<br>Keep experiments, edits, and reviews separate by default so parallel work does not turn the main branch into a collision zone.

Repo isolation<br>agent-a<br>worktree-1

agent-b<br>worktree-2

agent-c<br>worktree-3

$ isolated worktrees per agent<br>$ parallel edits<br>$ clean merge path

$ file editing<br>Edit repository files directly in the workspace.<br>Apply targeted edits where the work is happening instead of bouncing between disconnected tools. Nora keeps file editing in the same operational surface as the rest of the run.

Editing<br>$ src/app/page.tsx<br>find: "hero copy"<br>- legacy CTA text

$ in-workspace edits<br>$ targeted patches<br>$ faster iteration

*:first-child]:order-2 lg:[&>*:last-child]:order-1 translate-y-10 opacity-0" style="transition-delay:280ms">$ browser preview<br>Use the built-in browser to view what you're building.<br>Preview routes, verify UI behavior, and check rendered output without leaving Nora. The built-in browser keeps visual validation inside the same loop as agent execution.

Preview

$ diffs<br>Review code diffs before changes move forward.<br>Inspect exactly what changed, keep reviews explicit, and catch regressions earlier. Diff visibility is built into the workflow so operator approval stays grounded in code.

Review<br>$ git diff<br>src/components/feature-sections.tsx<br>- old hero paragraph

$ git diff visibility<br>$ change auditing<br>$ approval-ready output

*:first-child]:order-2 lg:[&>*:last-child]:order-1 translate-y-10 opacity-0" style="transition-delay:280ms">$ screenshots<br>Paste screenshots into terminal-based agents.<br>Bring visual context directly into CLI agent workflows. Nora lets you paste screenshots into terminal-based agents, which is not something they support natively on their own.

Visual context<br>$ paste screenshot

png

attached screenshot: login-state-bug.png<br>terminal agent can inspect UI regressions with visual context

$ paste screenshots<br>$ attach visual bugs<br>$ terminal agents with images

$ ide handoff<br>Hand work back to the IDE you already prefer.<br>Use Nora as the operating layer, then jump into your preferred editor for deeper inspection, local debugging, or final polish without losing the thread of the work.

Preferred editors<br>$ open in editor

current worktree

$ open current worktree in your IDE<br>$ keep your familiar editing flow<br>$ nora stays the control plane

*:first-child]:order-2 lg:[&>*:last-child]:order-1 translate-y-10 opacity-0" style="transition-delay:280ms">$ remote<br>Manage SSH-mounted repos like local projects.<br>Remote work should feel first-class. Nora is being shaped around SSH mount management so remote repositories can be operated on with the same clarity as local ones.

Remote mode<br>$ ssh mounts<br>prod-web-01mounted<br>staging-api-02ready

$ ssh mount management<br>$ remote terminals<br>$ remote agents

$ integrations<br>Connect the delivery layer, not just the editor.<br>Source control and deployment need to stay in the same flow. GitHub and Vercel integrations are part of the working surface, not an afterthought.

Connected stack

synced

$...