Agent Studio is a native desktop app for running, scheduling, and managing Claude-powered agent conversations.
Keep tool approvals, project context, skills, usage, and audit history in one focused workspace.
Available for macOS and Windows.
Built for real agent work on your computer
Agent Studio brings Claude agent conversations, tool calls, approvals, scheduled tasks, project context, skills, usage tracking, and audit history into one native desktop workspace.
Start chat-style agent runs that stream responses, reasoning, tool calls, and task history inside a focused desktop app.
Review tool calls before they run, choose permission behavior, and keep sensitive actions visible.
Create repeatable agent workflows for daily briefs, weekly reviews, inbox triage, and repository watches.
Shape how Agent Studio works with you through projects, instructions, skills, connectors, plugins, and agents.
Agent Studio is designed for knowledge workers who want Claude to do real work while keeping context, approvals, and history organized.
Start with Claude access, add project context, then run or schedule agent work from the desktop app.
Add your Anthropic API key or compatible endpoint, then choose the default model for new conversations.
Add instructions, memory, files, and context so the agent understands the work before it starts.
Start a task immediately, approve tool calls as needed, or turn repeatable prompts into scheduled workflows.
Track task history, usage, approvals, and audit events so agent work stays understandable after it finishes.
Built on the Claude Agent SDK, a Tauri desktop shell, and a Bun sidecar, Agent Studio keeps agent runs, approvals, schedules, and records in one local workspace.
Streaming responses, reasoning, tool calls, and replayable task history for long-running Claude agent work.
Approval cards for mutation tools, backed by workspace sandboxing, allowed tools, and destructive-command guardrails.
Run recurring work such as daily briefs, weekly reviews, inbox triage, and repository watches.
Group project instructions, memory, local folders, links, scheduled tasks, and related conversations.
Shape the app with personal skills, built-in skills, connectors, plugins, and agent configuration.
Track token and cost signals, tool activity, approvals, and redacted audit events with local event history.
No inflated customer counts or unsupported claims — just the facts Agent Studio is designed around.
desktop platforms: macOS and Windows
runtime layers: React UI, Tauri shell, Bun sidecar
workspace for tasks, projects, schedules, usage, and audit
Use Agent Studio wherever Claude needs context, tools, approvals, and repeatable execution.
Watch repositories, summarize failures, prepare maintenance notes, and keep review context in one project workspace.
Engineering teams
Repository workflows
Schedule daily briefs, weekly reviews, inbox triage, and status summaries without re-running the same prompt by hand.
Operators
Recurring briefs
Use projects, files, memory, and skills to keep multi-step research tasks grounded in the right context.
Researchers
Long-context work
Work with reports, spreadsheets, slide outlines, summaries, and other outputs while keeping the agent process visible.
Knowledge workers
Document workflows
Prototype permission-aware agent workflows with real tool calls, reusable skills, and auditable execution history.
AI builders
Agent prototyping
Turn repeated personal workflows into scheduled agent runs while keeping approvals and always-allowed tools under your control.
Power users
Personal automation
The essentials before you start using Agent Studio.
Run agent conversations, approve tool use, schedule recurring workflows, and keep project context organized.