Bring your agents. Keep your direction.

One Mac workspace for coding agents, shared plans and the work from first question to reviewed change.

Try the workflow Check availability

Bring the whole
workflow together.

A clear home for shared plans and the local work that turns them into changes.

Servlet Issues interface showing example planning data — A shared queue with clear ownership.
Plan in list or board views. Give each issue an owner, status, priority and labels, then discuss the work in comments.
Real product validation screen.
Example data. Earlier interface styling.

A WORKFLOW YOU CAN TRY

Turn a brief into
a reviewed change.

Explore the research-first path below. Servlet also supports a PRD-first workflow and direct sessions.

Choose an example

Interactive illustration.
No agent runs on this page.

THE BRIEF

Add keyboard navigation to the issue board without changing existing mouse behavior.

STAGE 1 OF 5 · Read only

Understand the problem first.

Inspect the project and gather evidence before deciding what to change.

Research documentEXAMPLE

Add a feature

Inspect the current implementation

Record the problem and supporting evidence

List assumptions and open questions

Work together.
Build on your Mac.

Shared planning gives the team a common view. Your coding workspace stays close to the files you work on.

Two precisely separated folded paper planes

Local coding workspace

Project folders and file paths

Coding tasks and session transcripts

Git worktrees, status and diffs

Coding-agent credentials with their CLIs

Your configured CLI may send context to its model provider. Local storage is separate from local inference.

Know what
the agent can do.

Choose the scope, inspect the request and keep delivery decisions in your hands.

Read-only stages

Research, requirements, design, planning and review use read-only access. Implementation follows the task's selected access.

Choose implementation access

Select the local project, agent, model, reasoning effort and task access before starting a session.

Exact pending approvals

Codex approvals are tied to the specific pending command or file request. Unsupported permission requests fail closed.

Review before publishing

Preview the exact text you choose to share. Local transcripts and documents are not automatically published to the team.

TRY AN APPROVAL

One request. Your decision.

Agent: Codex
Command: npm test
Scope: Example project folder

Waiting for your decision.

Illustration only. No command is executed.

A workspace with
room to think.

The everyday tools around an agent session matter just as much as the session itself.

Local terminal

Run a real native PTY in your project folder. Resize, clear or restart it without leaving the workspace.

Documents that follow the work

Edit Markdown and keep research, requirements, design, planning and review documents alongside your task.

Isolated Git worktrees

Give a task its own working directory, then inspect status and diffs. Archived worktrees remain available.

Sessions and saved drafts

Search conversations, save task drafts, interrupt a session and continue with its context intact.

Teams and organizations

Invite people, manage membership and access requests, and keep personal and organization workspaces separate.

A focused desktop

Use the command palette, keyboard shortcuts, local file selection, and Graphite or Paper themes.

Bring the agents you already use.

Choose Codex or Claude, an available model and reasoning effort. Stream the conversation, interrupt when needed, then continue.

CodexClaude

BEFORE YOU START

A few useful
answers.

Understand the workspace,
the boundaries and the next step.

Who is Servlet for?

Developers and engineering teams who already use coding agents and want one place for planning, sessions, context and reviewed delivery.

Which computers does it support?

The currently documented desktop build targets Apple silicon Macs running macOS 13 or later. The public installer is being prepared; check availability for release updates.

Do I need Codex or Claude installed?

Yes, for coding sessions. Servlet discovers your installed Codex and Claude CLIs and uses their existing configuration and account. Installation detection alone does not prove authentication or model access.

How does agent usage work?

Your selected CLI's configured provider and account determine model access, limits and usage costs. Servlet does not make every model available or include a separate model subscription.

What stays on my Mac?

Coding tasks, local project paths, transcripts and coding-agent credentials stay on the Mac. Accounts, organizations and shared planning records are hosted. Local coding workspaces do not synchronize between devices.

Is all agent processing local?

The coding workspace is local, but the selected CLI may send context to its configured model provider. Local storage does not imply local inference. Review your CLI configuration and provider terms.

Can a task run or advance by itself?

Creating a task saves its brief. You explicitly start the session and advance stages. Dependency readiness informs that decision; people still decide when shared work is Done.

How do approvals differ between agents?

Codex supports shell tools and exact pending command or file approvals. Claude supports confined file tools and displays denied permissions. Unsupported permission requests fail closed.

Can I download it today?

The public Mac installer is being prepared. The availability page is the source for public download status. Local validation builds are not offered here.

What happens to conflicting shared edits?

Shared records use version checks and actor-attributed activity. Conflicts are surfaced so you can review the latest version and preserve your draft rather than silently overwrite someone else's work.

Servlet for Mac. Public installer being prepared.

Check availability

Already installed? Open Servlet ↗Apple silicon · macOS 13+