Runtime CLI
Install once, then create agents, launch sessions, import runbooks, set guardrails, and read every run's grade from your terminal. Or hand it all to your coding agent.
Everything Runtime, from your shell
Everything Runtime, from your shell
Agents and triggers
List your agents, create new ones on a template, and connect them to Slack, Teams, email, or a schedule. Every command returns JSON for scripts and CI.
Sessions
Start a run on its own computer, stream what the agent is doing, and run any command inside the sandbox with clean JSON output.
Skills and runbooks
Write a runbook once as a skill. Runtime stores and versions it, every agent on the template picks up the change, and you can lock critical skills so only admins can edit them.
Guardrails and approvals
See every rule that applies to a template, from card-data hooks to network allowlists, and clear an approval without leaving the terminal.
Scheduled agents
Run sweeps on a schedule, like Radar reviews every 30 minutes or a daily reconciliation. Test once on demand, then turn it on.
Scorecards and cost
See how each agent performs, what it spends against its budget, and the grade on any single run.
Built for coding agents
Claude Code, Codex, and Cursor can run your whole Runtime setup: build the template, attach runbooks, wire the trigger, and prove it works before anyone turns it on.
JSON in, JSON out
Stdout is always JSON and errors come back as structured JSON with a hint, so scripts and agents parse results instead of scraping text.
Exit codes you can script
0 for success, 1 for an API error, 2 for auth, 3 for a usage mistake. Wire it into CI, cron, or your own tooling.
Ships with an agent skill
The installer adds a SKILL.md to Claude Code and Cursor, so your coding agent knows the golden paths and reads the docs before it acts.
CLI questions
How do I install the Runtime CLI?+
Run curl -fsSL https://runtm.com/install | bash. It installs a checksum-verified runtm-api binary for macOS or Linux, then set RUNTM_API_KEY from Settings > API Keys and run runtm-api auth status.
Can my coding agent use the CLI?+
Yes. runtm-api is built for coding agents like Claude Code, Codex, and Cursor. Output is JSON, exit codes are stable, and the bundled skill points the agent at docs.runtm.com/llms.txt for every command.
What is the difference between runtm and runtm-api?+
runtm is the Python CLI for local sandboxes, scaffolding, and deploys. runtm-api is a Go binary that drives the hosted Runtime platform: agents, sessions, skills, guardrails, schedules, and scorecards. The same API key works for both.
Which commands need an org-scoped API key?+
Templates, agents, scheduled agents, skills, MCP servers, tools, guardrails, team secrets, and org instructions need an org-scoped key. Writes to agents, triggers, and guardrails also need an admin or owner role.
Build your team of agents
Start from the terminal or the dashboard. Same agents, same guardrails, same audit trail.