Skribly Agent

Overview

Skribly Agent is a desktop environment for AI coding agents. It keeps your projects, tasks and agents in one window, so you do not jump between terminals.

The app reads each agent's screen and shows whether it is working, waiting for your decision or finished. When something needs you, it tells you, even if you are looking elsewhere.

Agents side by side

Claude Code, Codex, Cursor Agent, Gemini CLI and others in workspaces, tabs and split panes.

Tasks from trackers

Jira, Redmine, GitHub, GitLab and Linear in one list. Start a task and the agent gets a branch, a file and a prompt.

Git worktrees

Each agent can work in its own copy of the files, so agents do not get in each other's way.

Pipelines

Plan, build and review steps that hand work from one agent to the next.

Audits

Security, accessibility and SEO checks, with results you can pass to an agent.

A real editor

Monaco with tabs, search in files, side-by-side diffs and a history graph of all branches.

Commits in your style

Suggestions from your repository's history, without a model and without AI trailers.

Browser and Docker

Preview at phone, tablet and laptop sizes, and a view of your containers.

CLI and MCP

Agents can start, prompt, wait for and read other agents.

Accessibility settings

Text size, high contrast, reduced motion, underlined links and a strong focus indicator.

Themes and languages

Day and night modes, editor themes, and an interface in Polish and English.

Workflows

From a ticket to a commit

1. Open Tasks and choose Start with an agent.
2. The agent reads TASK.md, proposes a plan and waits for your OK.
3. It works in stages and updates STATE.md after each one.
4. You read the changes in Code & git.
5. Suggest writes the commit message, and you commit.

A reviewer next to the implementer

One agent starts another through the skribly command, waits for it and reads its comments. You can watch the reviewer in its own pane and reply to it.

skribly agent start reviewer --kind claude --prompt "Review the changes from git diff."
skribly agent wait reviewer --until done --timeout 600
skribly agent read reviewer --lines 60 --text

Claude plans, Codex builds, Claude reviews

Plan, implement and review is one of three built-in pipelines. The reviewer checks git diff and suggests a commit message.

Three audits at once

Security, accessibility and SEO run in parallel with three agents, then a coordinator merges the results into one plan of fixes.

Move a session to another agent

The current agent saves its state to a file, and the new one starts next to it in the same worktree, without replaying the conversation.

From a design to code

Export a package from Claude Design, choose an agent and choose Implement. It proposes a plan, then builds from your project's components.

Open larger image: Claude Design to code

Read the CLI and MCP reference

Shortcuts

Ctrl+B works as in tmux: press it, release it, then press the key. On Windows and Linux, use Ctrl instead of ⌘.

See every shortcut

Security

• No account, no telemetry and no server of ours. Agents use their own logins and API keys.
• Tracker tokens are encrypted with the system password store.
• The control API listens only on 127.0.0.1 and needs a token that changes on every startup.
• Browser sharing is off by default. When it is on, any program on the computer can reach it, so use it only while you work with agents.
• Agent terminals run with your permissions.

See what is stored and where it goes

Details

Screenshots

Open larger image: Dashboard

Open larger image: Agents

Open larger image: Tasks

Open larger image: Code changes

Open larger image: Pipelines

Open larger image: Themes

Open larger image: Dashboard in the day theme

Open larger image: History graph

Open larger image: Audits

Open larger image: MCP servers

Open larger image: Claude Code plugins

Open larger image: Claude Design to code

Open larger image: Settings

Open larger image: Built-in tour

Recordings