reachpad: run and share the software your coding agents build<br>The software your agent writes is already running.<br>Point Claude Code, Codex or OpenCode at a workspace. What it builds runs there, keeps its data and its secrets, and one command turns it into a link you can send.<br>Start trialSet up MCP<br>HomebrewInstall scriptnpmMCP<br>$brew install reachpad/tap/reachpadcopy
Preinstalled in every workspace<br>Claude CodeCodexOpenCode
The software that was never worth deploying.<br>Agents made all four of these cheap to write. The cost was always everything after that: somewhere to run, and a way for the person who needs it to open it.
What your agent builds<br>What the person who needs it gets
A dashboard over the support CSV<br>it would have stayed a notebook
A page they open themselves, reading the same file
A tool that reconciles the expense export<br>nobody was ever going to deploy this
Still at the same address next month
A screen over a script only you can run<br>the runbook says to ask you
A control somebody else can use without you
A prototype that turned out to matter<br>it only ever ran on your machine
Running where the people who need it can reach it
Use the agents and workflows you already have.<br>A model calling tools, your own shell, or code you wrote. Whichever one builds it, the software lands in the same place.
create_environmentokrun_commandexit 0checkpoint_environmentsealed<br>MCPhttps://reachpad.dev/mcp<br>Claude and ChatGPT<br>Add one connector and the model makes the software and the place it runs in one move.<br>Connect a model→
$reachpad create$reachpad run -- npm start$reachpad fork<br>CLIbrew install reachpad<br>Your terminal<br>Start a workspace, run what is in it, fork it to try a second version.<br>Install the CLI→
create201exec200fork201<br>RESTPOST /v1/environments<br>Your own agent<br>The same workspaces over HTTP, from code you write.<br>Read the API docs→
Log inStart trial