connect your agent
claude code and cursor can run your pushto account. one line, no tokens to copy.
the one-liner
log in once with the cli, then hand your agent the keys. the easiest way is to let the cli do it:
pushto mcp add
it finds claude code or cursor and wires the config for you. doing it by hand is one line in claude code:
claude mcp add pushto -- pushto mcp serve
or one entry in cursor's mcp config:
{
"mcpServers": {
"pushto": { "command": "pushto", "args": ["mcp", "serve"] }
}
}auth rides your cli login. the server mints itself short-lived tokens behind the scenes; nothing to paste, nothing to store.
agents without the cli
running an agent on a hosted runner, a server, or anywhere the cli isn't? the same tools are served over http. mint a token in the dashboard (settings → connected agents), then call:
POST https://pushto.host/api/mcp Authorization: Bearer <your token>
GET on the same url lists every tool with its json schema. tokens expire after 7 days, and "log out all agents" kills every outstanding one instantly.
what your agent can do
the full loop: create a project, deploy the folder it is working in, read the runtime logs and the plain-english error episodes when something breaks, fix it, deploy again. plus the setup verbs: database, secrets, domains, auth, mail, legal, launch review.
| tool | what it does |
|---|---|
| pushto.deploy.local | deploy the folder the agent is working in. no repo connection needed. |
| pushto.project.describe | project status + service overview |
| pushto.project.create | create a new project |
| pushto.deploy.connect_repo | connect a github repo |
| pushto.deploy.detect_framework | detect the project framework |
| pushto.deploy.create | trigger a deploy of pushed code |
| pushto.deploy.status | poll one deployment |
| pushto.db.provision | provision the managed postgres |
| pushto.auth.configure | enable the drop-in login widget |
| pushto.domain.add | attach a custom domain |
| pushto.dns.verify | check custom domain dns + ssl |
| pushto.secrets.set | set one encrypted env var |
| pushto.secrets.list | list env var names |
| pushto.mail.configure | set up domain email |
| pushto.legal.generate | generate privacy policy + terms |
| pushto.logs.tail | tail the runtime logs |
| pushto.errors.episodes | read classified error episodes |
| pushto.launch_review.run | start the pre-launch review |
| pushto.launch_review.latest | latest pre-launch review results |
this list renders from the live tool registry, so it is always current.
logging agents out
agent tokens expire on their own after 7 days. to kill every outstanding token right now, use the log out all agents button in your dashboard settings. machines where the cli is still logged in reconnect on their own; logging the cli out on a machine cuts it off completely.