- All machines
- This Computer
- Paired computers and SSH hosts
- Online and offline state
- Jump to Machines settings
Run a team of coding agents
The only dev tool you need in the AI era.
Brief a team, not a chat. Claude Code, Codex, Antigravity and the rest work the same job in parallel — one window, one live map of who is doing what. Click any pin below to see that feature, or take the guided tour through all 14.
One real screenshot
100%
No fork. No wrapper. No proxy. It runs the same command you'd type yourself.
Works withClaude CodeCodexAntigravityGrokGemini CLICursorKimiQwenOpenCodePi+ any CLI
Trusted by developers shipping with AI daily →49,805
Downloads
100+
Daily active devs
133
Releases · updated weekly
7-day refundNo subscription evermacOS · Windows · Linux
The problem
Stop being the human middleware.
Five agents, four terminal tabs, three projects, and you in the middle: copy-pasting context, checking who is stuck, re-typing prompts after every rate limit.
Your agents have no way to talk to each other, so you are the integration layer. 1DevTool gives them a chart instead of you.
The same afternoon, in one window
1You are the integration layer
Every agent talks to you. Every tool talks to you. Nothing talks to anything else.
❯You
claude codedirector
codexbackend
geminireviewbrowser
database
HTTP
git
An org chart or a pipeline for your terminals: who reports to whom, who hands work to whom. How orchestration works →
In the app
Watch one brief run the whole team.
The chart plays before a single terminal opens, and Start stays one deliberate click.
Every machine you code on — this laptop, a paired desktop, a server over SSH, and the terminals 1DevTool never started.
- Pairing
- Prompting a peer's agent
- Per-peer permissions
- Networking
- Nothing runs until you pair something
- Licensing
- Turning It On
- What You Can Do From Your Phone
- Notifications
- Permissions
- What are SSH Keys?
- View Existing SSH Keys
- Generate a New Key
- Add to GitHub or GitLab
- Use an SSH Key for Git
- Connect to a Remote Server
58 projects grouped the way you think. Switch and the panes, branch and worktree come back exactly as you left them.
- Color-Coded Project Icons
- Right-Click Menu
- Drag to Reorder
- Project Groups
- Active Agent Badge
- System Tray
- Creating one
- Mentioning another project
- Agents can address the whole product
- Tasks across the workspace
- Where a workspace shows up
- The .1devtool/ folder
- Editing it
- Startup commands
- Tasks travel with the branch
- Sharing vs keeping it local
- What does not go in it
- Sizing the tree
- Collapsing it
- A standalone File Explorer
- Coloured file icons
- Search
- Expansion state that survives
- What is a Project?
- Add a Local Folder
- Clone from Git
- Add a Remote Server via SSH
Every coding CLI you run, as a first-class object: named, coloured, pinned to a model, and resumable.
- Register a command as an agent
- Name, colour and logo
- Screen mode
- Sessions
- Memory files
- Rules and limits
- What Session Persistence Means
- How It Works: tmux
- No Configuration Needed
- The Session Manager
- Reconnecting After a Restart
- Opening it
- What you see per session
- Supported agents
- Resuming a whole team
- Already-open sessions
- Changing agent on resume
- Idle agents hibernate
- Hibernating on purpose
- Browser tabs unload too
- When memory runs short
Open a terminal for any agent on this machine, name it, and pin the model it launches with — one tab per agent, added in a click.
- Two ways to open the dialog
- What you can open
- Naming
- Opening several at once
- Startup commands
- Choosing which account it starts on
- Choosing an Agent
- Starting an Agent
- Typing a Prompt
- The Agent Input composer
- Knowing how full the context is
- Live run timers
Tile them however the work wants — grid, columns, mosaic, office, vibe. Switching layout never rebuilds a pane.
- Everything is a tile
- Splitting, stacking, resizing
- Several arrangements per project
- Keyboard
- Agents as characters
- Review inbox
- Sprint wall
- Levels, hats, and pets
- Decorating and sharing
- The weekly recap
- Vibe Home
- Submit, and it becomes a conversation
- Preview, Changes, and Code
- When to use it
A genuine PTY running the command you'd type yourself. No fork, no wrapper, no proxy.
- Terminal Recording
- Terminal Color Themes
- Reader Mode
- Send File Path to Agent
- Combine Sessions
- Terminal Fonts
- Three buckets that add up
- The attention strip
- Machine
- Recent stalls
- Copy diagnostic report
- Improve performance
- The window stopped responding
- Diagnostics settings
- The app is using too much memory
- One terminal is pinning the CPU
- Terminals feel slower than they should
Terminal output is writing. Read it like a page instead of scrolling a log.
- Conversation ⇄ Terminal toggle
- Answer multiple-choice questions with a click
- The agent needs your answer banner
- Resizable composer, remembered per terminal
- Load earlier for long transcripts
- Opening it
- What it fixes
- A terminal beside it, in a real split
- Prompting while you read
- Notes in Reader Mode
- On your phone
- Where notes live
- Markdown, written and read
- The notes library
- Typing in other languages
- On your phone
- Performance
- Split View
- What Gets Rendered
- Sticky Notes
- Common Use Cases
- What is Mermaid?
- Live Diagram Rendering
- Interactive Editing
- Use with AI Agents
- Browsing the File Tree
- Working with Tabs
- Quick File Open
- Drag File Paths into Terminals
- Tips
- Describe it and the canvas draws it
- Undo is one step
- Choosing the agent
- Draw or Mermaid?
A proper editor that types into the terminal — markdown, mentions, commands, personas, scheduled sends.
- Floating and resizable
- Drafts are never lost
- Mentions
- Choosing an orchestration pattern
- Attachments
- A verified send
- Building the library
- Turning them on per terminal
- Inserting one at the cursor
- What to put in one
- Templates vs skills vs project instructions
- Opening the browser
- Editing
- Inserting one into a prompt
- From your phone
- One library, every project
- Orchestration skills
- The library
- Pinning and applying
- Sharing a skill set with your team
- Rolling forward and back
One brief fans out, every agent works at once, this terminal merges the answers.
- The shapes
- Drawing one
- Picking the right one
- Where you watch it happen
- Delivery you can trust
- Permissions are opt-in
- Teams: agents you can watch
- Swarms: headless workers
- When a run fails
- Native protocols where they exist
- Resuming a team
- Seeing teams from anywhere
- The chart is enforced, not decorative
- Repairing a team without rebuilding it
- Refusals are visible
- Zero-setup chains
- From your phone
- Stages
- Quality gates
- Starting one from the composer
- Watching it run
- Pipeline or hierarchy?
- Nothing is readable until you say so
- Revoking
- Proving who is asking
- Publishing to a peer
- The link list in Settings
- A note on trust
The live map of who is linked to whom, who is working, and who is waiting on you.
- Opening it
- The participant rail
- Counters you can act on
- Recovery in one action
- When a prompt did not send
- Approvals on the map
- Where to find it
- Reading a card
- Reading an edge
- Linking by dragging
- Zoom, re-center, and remembered positions
- The activity rail
- Mention a terminal
- Get the answer back
- Delivered, or delivery-unconfirmed
- Approve a delivery before it lands
- Terminals arrange themselves
- When a terminal restarts
- Teams
- Team Map
- Routing
- Run & Logs
- Usage
- Skill
- What Is a Channel?
- Why Use Channels?
- How to Open Channels
- A Typical Workflow
- Choose an Agent and Model
- Choose a Reasoning Level
- Review and Refresh Available Models
- Verify the Handoff
- Pick the Right Model for the Task
Sessions, prompts, memories, skills, MCP, diffs and tasks — the whole workbench in one row.
- Tools, grouped by what they touch
- Turning a tool off
- When an agent hits a wall
- Connecting agents
- What an agent can do
- How it stays under control
- Big pages and screenshots
- Asking for it explicitly
- Turning it off
- Opening Tasks
- The views rail
- Tasks are files in your repo
- Cards say who is on it
- Nothing auto-approves
- Agents work them over MCP
- Assigning
- Agents cannot assign themselves
- Watching it run
- Unassigning
- Queueing behind a busy agent
- Dragging is not assigning
- The two gates
- Answering from anywhere
- Nothing auto-approves
- Gates and the board
- The Needs me view
- Capture from where the problem appears
- Decompose a goal
- Writing one by hand
- Merging duplicates
- Opening it
- What you can do
- Picking a schedule
- Reading the logs
- Popping it out
- Runtime tools
The apps you used to alt-tab to, docked beside the terminal that needs them — and reachable by your agents.
- Why It Matters
- What the Embedded Browser Supports
- How to Open the Browser Panel
- Positioning in Your Layout
- What a profile is
- Create and manage profiles
- Switch a tab's login from the toolbar
- Agents inherit the profile
- Tell sites how the browser should look
- Browse and clear a profile's history
- What Is the Database Client?
- Supported Database Engines
- How to Open the Database Client
- What You Can Do
- The Query Editor
- Running a Query
- Autocomplete
- Results Grid
- Row Detail Drawer
- Query History
- What the Schema Browser Shows
- Exploring a Table's Structure
- Quick Data Preview
- Hovering on Columns
- Searching for a Table
- Refreshing the Schema
- What Is the HTTP Client?
- Why Use It Instead of Postman?
- How to Open the HTTP Client
- Key Features
- Who Is This For?
- Your First Request
- Add Request Headers
- Add a Request Body
- Inspect the Response
- Save a Request
- Opening it
- What is in it
- The password generator
- Everything runs locally
Branch and worktree in the status bar; the full client one keystroke away.
- What Is the Git Client?
- Why Use It?
- How to Open
- What You Can Do
- The status-bar branch picker
- Actions on a commit
- Opening the Diff View
- Reading the Diff
- Unified vs Side-by-Side View
- Image Diffs
- Keyboard Navigation
- Why Review Before You Commit?
- What is a Git Worktree?
- Create a Worktree
- Worktree Terminals
- Switch Worktrees From the Status Bar
- Remove a Worktree
- Adding a second root
- The repository picker
- Change badges in the explorer
- Repo-scoped worktrees
- Terminals and worktrees
- Git accounts per repository
What your agents are costing you — in subscription quota and in machine.
- The Usage Pill
- Per-Agent Quota
- The Spend & Quota Center
- Switching Accounts to Stretch Your Quota
- Usage on Your Phone
- Three buckets that add up
- The attention strip
- Machine
- Recent stalls
- Copy diagnostic report
- Improve performance
- What free gives you
- What Pro unlocks
- When you hit a limit
- Activating a licence
- The suite bundle
Compared, honestly.
Everyone runs parallel agents in 2026. The difference is what surrounds them: failover, drawn charts, a full workbench, your own subscriptions, and a price that is not a meter.
Feature reading as of August 2026. First-party CLIs run their own vendor's models; Conductor is macOS-only; Warp and Cursor meter usage through credit plans.
And it replaces the single-purpose tools.
tmux, VS Code, Postman, TablePlus, Sublime Text: every tool in your dev stack lives in its own window. 1DevTool fuses them into one workspace with AI built in.
5 separate apps → 1 workspace. $29 once. No subscriptions.
Stack cost over 3 years: ~$700 with TablePlus, Sublime Text, and Postman Basic bought separately (prices as of Aug 2026).
Most devs glue together a stack of single-purpose apps: tmux + iTerm for terminals, VS Code or Sublime Text for editing, Postman for HTTP, TablePlus for databases. 1DevTool combines all five into one workspace, with parallel AI CLIs, rich prompts, and cross-tool context built in, for a one-time $29. No subscriptions, no per-tool seats.
Developers who switched
Real users, real names, linked to their sites.
"1DevTool has gotten really impressive. The UX redesign feels much more intuitive. I use the app every day, and seeing consistent updates like this gives me real confidence in it."
"Running Claude Code + Codex across 3-4 projects used to mean a desktop of terminal tabs. 1DevTool puts every project (terminals, browser, HTTP client, DB) in one window, and remembers exactly where each agent left off when I close the app. $29 once, replaces my whole AI dev stack. Best price-to-performance in the market."
"As a fullstack software developer, I use AI to assist with coding almost constantly, every single day — choosing to use and pay for 1DevTool has been the perfect choice for my work. I've been using 1DevTool since day one, and I really appreciate the support as well as the guidance on how to use the app. The developer has always strived to improve and update 1DevTool in the best possible way, based on users' needs. I've also just paid for the 1FileTool app to support my work. Thank you so much, 1DevTool."
"I work with multiple AI agents and projects simultaneously every day. 1DevTool is the first workspace that makes this setup actually manageable. Being able to run and monitor many terminals in parallel while keeping each project organized in its own persistent workspace saves me a huge amount of time and mental overhead."
Pricing
Free to start. $29 once. No meter.
1DevTool never sits between you and your AI: your Claude, OpenAI, and Google subscriptions stay yours, with no credits and no markup. Pay once for the workbench when you need more room.
Free
$0forever
No account required
- 1 project · 4 terminals · 1 browser tab
- Unlimited prompts to AI terminals
- Agent @mention orchestration
- 1 DB connection · 5 saved HTTP requests
- 5 AI diffs/day · 7-day prompt history & memory
Most popular
Pro
$29one-time · 1 device
3 devices $59 · 5 devices $89 · 12 months of updates
- Unlimited projects, terminals, browser tabs
- Unlimited DB connections, HTTP requests, git worktrees
- Unlimited AI diffs, prompt history, resume & memory
- Project Settings, project groups, custom themes & shortcuts
- 7-day money-back guarantee
Secure checkout via Lemon Squeezy · Bring your own AI CLI, we never charge for AI usage
Frequently asked questions
What AI agents does it support?
Claude Code, Codex, Gemini CLI, Cursor, Grok, Kimi Code, Antigravity, Hermes, Pi, Amp, OpenCode, Cline, Qwen Code, and Aider, plus bash, zsh, and PowerShell. You can define custom commands too: if it runs in a terminal, it works in 1DevTool.
How is this different from Warp, Conductor, or Cursor's parallel agents?
They all run agents in parallel now; the difference is what surrounds them. 1DevTool runs the first-party CLIs locally on your own subscriptions (no credits, no markup), lets you draw org charts and pipelines for how agents hand work to each other, fails a dead seat over to a backup agent automatically, and bundles the editor, browser, HTTP client, database client, and phone remote in the same window. It runs on macOS, Windows, and Linux for a one-time $29.
What happens when an agent hits a rate limit or crashes?
Two safety nets. Per seat: any seat on a chart can name up to three backup agents; when the primary's process dies (a crash, an out-of-quota exit), the next backup opens in that seat and receives the conversation so far. Per type of work: fallback chains route unnamed work to the next engine when a run fails, and Active Teams shows the hop that happened. Closing a tab or quitting the app never counts as a failure.
Can I control my agents from my phone?
Yes. Pair your phone by scanning a QR code, over your local network or from anywhere via a secure Cloudflare Tunnel (works on LTE). You can watch terminals live, send prompts with @ file mentions and image attachments, approve permission requests, browse and edit files, use the git client, run HTTP requests and database queries, and get push notifications the moment an agent finishes.
Does it work on Windows and Linux?
Yes. macOS (Apple Silicon + Intel), Windows 10/11 (NSIS installer + portable exe), and Linux (AppImage + .deb). Full PTY terminal support on all platforms. Tmux-based session persistence is available on macOS and Linux.
What happens after I buy? Is there a subscription?
No subscription, one-time payment. Your license includes 1 year of free updates. After that, the app keeps working forever, you just won't receive new feature updates unless you renew. Licenses are per-device (1, 3, or 5 devices depending on your plan) and you can deactivate and move them anytime.
Can I use it with local AI models (Ollama, LM Studio)?
If the model has a CLI interface, you can run it as a terminal in 1DevTool. There's no built-in integration with Ollama or LM Studio specifically, but any command-line tool works as a custom agent.
What databases does it support?
23 database engines across SQL (PostgreSQL, MySQL, MariaDB, SQL Server, SQLite, CockroachDB, ClickHouse, TimescaleDB, Hydra, Neon, Supabase, Amazon Redshift), NoSQL (MongoDB, Elasticsearch, CouchDB, Cassandra), cache (Redis, Valkey, KeyDB), vector/multi-model (SurrealDB, Weaviate), and specialized (InfluxDB, Apache Kafka). Each has schema browsing, query execution, and export support.
What are Agent Pipelines?
Pipelines let you chain AI agents together across multiple terminals. You can set up conditional branching (regex match, contains), delay and checkpoint nodes for human approval, output capture modes (full, regex, last N lines), and transform operations between steps. Great for multi-step workflows like Plan → Review → Revise with different AI agents at each stage.
Is there a free tier?
Yes, a permanent free tier with no time limit. You get unlimited prompts to AI terminals, Agent @mention orchestration, 1 project, 4 terminals, 1 browser tab, 1 git worktree, 1 database connection, 5 saved HTTP requests, 5 AI diffs per day, and 7-day prompt history, resume & memory. All platforms supported. No credit card required.
What's the refund policy?
7-day money-back guarantee, no questions asked. Processed through Lemon Squeezy. If 1DevTool isn't right for you, contact us for a full refund.
Does 1DevTool support MCP?
Yes, 1DevTool ships a built-in MCP server that exposes your workspace to any MCP-capable agent. Your agents can query your saved database connections (query, preview_table, describe_schema), run your saved HTTP requests with environments, and more, with live MCP tool-call badges on the terminal so you always see what the agent touched.
How does Send to AI work?
The embedded browser can capture screenshots (including lazy-loaded content and JS state), annotate them with a built-in markup tool, and send the annotated screenshot plus URL context and console logs directly to any AI agent running in your terminals, all in one click.
I already use tmux / iTerm2 splits. Why switch?
1DevTool adds project-level persistence with named agent sessions, a file tree, a built-in database client supporting 23 engines, an HTTP client, an embedded browser with screenshot annotation, Send-to-AI integration, agent orchestration, and smart notifications. tmux doesn't know what project you're in; 1DevTool does.
Your agents are already parallel.
Your workspace isn't.
Run every coding agent like a team: drawn charts, automatic failover, one window that sees your editor, browser, HTTP client, and database. And a phone in your pocket that sees all of it too.
Free forever tier · $29 one-time to unlock everything · No subscription
Multi-CLI terminalsOrg charts & pipelinesBackup agentsConversation modeAI Diff reviewPhone remote controlSession resumeSkills & MCPEmbedded browserHTTP client23-engine DBGit worktrees